From c7a29ec0a7890e4f8217687cf04fc9582ba9b1f4 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 12:51:44 +0800 Subject: [PATCH 01/24] chore(storage): drop dead backpressure and lock-optimizer wrappers (#6195) Neither rustfs/src/storage/backpressure.rs nor rustfs/src/storage/lock_optimizer.rs had a production caller: their only non-self references were the pub mod lines in storage/mod.rs and a cfg(test) module, so the object transfer path never applied this backpressure and never took these lock shortcuts. The six removed tests in concurrent_fix_test.rs duplicated tests that lived inside the deleted files; the shared primitives they shadowed keep their own coverage in rustfs-io-core. --- crates/concurrency/src/backpressure.rs | 5 +- docs/architecture/global-state-inventory.md | 1 - rustfs/src/storage/backpressure.rs | 618 -------------------- rustfs/src/storage/concurrent_fix_test.rs | 82 +-- rustfs/src/storage/lock_optimizer.rs | 458 --------------- rustfs/src/storage/mod.rs | 2 - 6 files changed, 4 insertions(+), 1162 deletions(-) delete mode 100644 rustfs/src/storage/backpressure.rs delete mode 100644 rustfs/src/storage/lock_optimizer.rs diff --git a/crates/concurrency/src/backpressure.rs b/crates/concurrency/src/backpressure.rs index 502d4fb0f..e692d7329 100644 --- a/crates/concurrency/src/backpressure.rs +++ b/crates/concurrency/src/backpressure.rs @@ -14,9 +14,8 @@ //! Shared backpressure policy type. //! -//! The runtime backpressure implementation (byte-watermark pipes and -//! monitors) lives in `rustfs/src/storage/backpressure.rs`; this module only -//! carries the watermark policy type that implementation shares. +//! This module only carries the watermark policy; the admission primitive it +//! projects into lives in `rustfs-io-core`. use rustfs_io_core::BackpressureConfig as CoreBackpressureConfig; diff --git a/docs/architecture/global-state-inventory.md b/docs/architecture/global-state-inventory.md index 8407ad864..b110374dd 100644 --- a/docs/architecture/global-state-inventory.md +++ b/docs/architecture/global-state-inventory.md @@ -105,7 +105,6 @@ inventory. Generic function-local names such as `CACHE`, `LOCK`, `INIT`, and | `USE_STARSHARD_CACHE`, `BUCKET_CACHE_SMALL`, `BUCKET_CACHE_LARGE` | `rustfs/src/storage/ecfs_extend.rs` | Cache or constant / owner-local cache | Bucket validation cache backend selection and cache storage stay private to the ECFS extension owner. | | `GLOBAL_SSE_DEK_PROVIDER`, `SSE_TEST_LOCK` | `rustfs/src/storage/sse.rs` | Owner-local cache / test state | SSE DEK provider cache and test serialization lock stay private to the SSE owner. | | `AUTH_FS` | `rustfs/src/storage/access.rs` | Cache or constant / owner-local cache | Authorization tag-condition lookup keeps its filesystem helper private to the access owner. | -| `LOCK_STATS` | `rustfs/src/storage/lock_optimizer.rs` | Process-global owner-local metrics | Lock optimization statistics stay private behind lock optimizer helper APIs. | | `DEADLOCK_DETECTOR` | `rustfs/src/storage/deadlock_detector.rs` | Process-global owner-local state | Deadlock detector lifecycle state stays private to the storage deadlock detector owner. | | `CONCURRENCY_MANAGER`, `ACTIVE_GET_REQUESTS`, `ACTIVE_PUT_REQUESTS` | `rustfs/src/storage/concurrency/*` | Process-global owner-local scheduler state | Storage concurrency manager and request counters remain inside the storage concurrency owner boundary. | | `GET_OBJECT_BUFFER_THRESHOLD_WARNED`, `GET_READER_STREAM_BUFFER_SIZE_OVERRIDE`, function-local `ENABLED`, `OBJECT_SEEK_SUPPORT_THRESHOLD`, `OBJECT_SEEK_SUPPORT_CONCURRENCY_THRESHOLDS` | `rustfs/src/app/object_usecase.rs` | Cache or constant / owner-local cache | Object GET/seek tuning caches and warning guards stay private to object usecase helpers. | diff --git a/rustfs/src/storage/backpressure.rs b/rustfs/src/storage/backpressure.rs deleted file mode 100644 index 9585ba8a8..000000000 --- a/rustfs/src/storage/backpressure.rs +++ /dev/null @@ -1,618 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Backpressure Management for Object Data Transfer. -//! -//! This module provides backpressure-aware pipes for object data transfer, -//! preventing buffer overflow and memory exhaustion under high concurrency. - -//! # Key Features -//! -//! - Configurable buffer size with high/low watermarks -//! - Backpressure state monitoring and events -//! - Backpressure metrics emitted through the shared metrics pipeline -//! - Graceful handling of slow consumers -//! -//! # Architecture -//! -//! ```text -//! [Disk Reader] --> [BackpressurePipe] --> [HTTP Response] -//! | -//! v -//! [Buffer Monitor] -//! | -//! v -//! [High Watermark?] --> Apply Backpressure -//! ``` - -use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering}; -use std::time::{Duration, Instant}; -use tokio::io::{DuplexStream, duplex}; -use tracing::{debug, warn}; - -use metrics::counter; -use rustfs_concurrency::PipeBackpressurePolicy; -use rustfs_io_core::BackpressureConfig as CoreBackpressureConfig; - -/// Object-transfer duplex pipe backpressure policy. -#[derive(Debug, Clone, Copy)] -pub struct ObjectPipeBackpressurePolicy { - /// Buffer size in bytes (default 4MB). - pub buffer_size: usize, - /// High watermark percentage (default 80%). - /// When buffer usage exceeds this, backpressure is applied. - pub high_watermark: u32, - /// Low watermark percentage (default 50%). - /// When buffer usage drops below this after high watermark, backpressure is released. - pub low_watermark: u32, -} - -impl Default for ObjectPipeBackpressurePolicy { - fn default() -> Self { - Self { - buffer_size: rustfs_config::DEFAULT_OBJECT_DUPLEX_BUFFER_SIZE, - high_watermark: rustfs_config::DEFAULT_OBJECT_BACKPRESSURE_HIGH_WATERMARK, - low_watermark: rustfs_config::DEFAULT_OBJECT_BACKPRESSURE_LOW_WATERMARK, - } - } -} - -impl ObjectPipeBackpressurePolicy { - /// Load configuration from environment variables. - pub fn from_env() -> Self { - let buffer_size = rustfs_utils::get_env_usize( - rustfs_config::ENV_OBJECT_DUPLEX_BUFFER_SIZE, - rustfs_config::DEFAULT_OBJECT_DUPLEX_BUFFER_SIZE, - ); - let high_watermark = rustfs_utils::get_env_u32( - rustfs_config::ENV_OBJECT_BACKPRESSURE_HIGH_WATERMARK, - rustfs_config::DEFAULT_OBJECT_BACKPRESSURE_HIGH_WATERMARK, - ); - let low_watermark = rustfs_utils::get_env_u32( - rustfs_config::ENV_OBJECT_BACKPRESSURE_LOW_WATERMARK, - rustfs_config::DEFAULT_OBJECT_BACKPRESSURE_LOW_WATERMARK, - ); - - Self { - buffer_size, - high_watermark, - low_watermark, - } - } - - /// Calculate high watermark threshold in bytes. - pub fn high_watermark_bytes(&self) -> usize { - (self.buffer_size as u64 * self.high_watermark as u64 / 100) as usize - } - - /// Calculate low watermark threshold in bytes. - pub fn low_watermark_bytes(&self) -> usize { - (self.buffer_size as u64 * self.low_watermark as u64 / 100) as usize - } - - /// Project this object-transfer policy into the shared concurrency facade policy. - pub fn to_concurrency_policy(&self) -> PipeBackpressurePolicy { - PipeBackpressurePolicy { - buffer_size: self.buffer_size, - high_watermark: self.high_watermark, - low_watermark: self.low_watermark, - } - } - - /// Project this object-transfer policy into the reusable io-core admission config. - pub fn to_core_config(&self) -> CoreBackpressureConfig { - self.to_concurrency_policy().to_core_config() - } -} - -/// Backpressure state. -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum BackpressureState { - /// Normal operation, buffer usage is below high watermark. - Normal, - /// Buffer usage is above high watermark, backpressure should be applied. - HighWatermark, - /// Backpressure is actively being applied to the producer. - BackpressureApplied, -} - -impl std::fmt::Display for BackpressureState { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - match self { - BackpressureState::Normal => write!(f, "normal"), - BackpressureState::HighWatermark => write!(f, "high_watermark"), - BackpressureState::BackpressureApplied => write!(f, "backpressure_applied"), - } - } -} - -/// Compact metadata snapshot for object-transfer backpressure pipes. -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub struct BackpressurePipeMeta { - /// Buffer capacity in bytes. - pub buffer_capacity: usize, - /// Current backpressure state. - pub state: BackpressureState, - /// Age of the pipe since creation. - pub age: Duration, -} - -/// Compact metadata snapshot for the lightweight backpressure monitor. -#[derive(Debug, Clone, Copy, PartialEq)] -pub struct BackpressureMonitorMeta { - /// Buffer capacity in bytes. - pub buffer_capacity: usize, - /// Current buffer usage percentage. - pub usage_percent: f32, - /// Current backpressure state. - pub state: BackpressureState, -} - -fn calculate_usage_percent(usage: usize, capacity: usize) -> f32 { - if capacity > 0 { - (usage as f32 / capacity as f32) * 100.0 - } else { - 0.0 - } -} - -fn apply_watermark_transition( - in_high_watermark: &AtomicBool, - usage: usize, - high: usize, - low: usize, -) -> (BackpressureState, bool) { - let current = in_high_watermark.load(Ordering::Acquire); - let next_state = if usage >= high { - BackpressureState::HighWatermark - } else if usage <= low { - BackpressureState::Normal - } else if current { - BackpressureState::HighWatermark - } else { - BackpressureState::Normal - }; - let next_is_high = matches!(next_state, BackpressureState::HighWatermark); - let changed = in_high_watermark.swap(next_is_high, Ordering::AcqRel) != next_is_high; - (next_state, changed) -} - -fn saturating_sub_atomic(value: &AtomicUsize, delta: usize) { - value - .fetch_update(Ordering::AcqRel, Ordering::Acquire, |current| Some(current.saturating_sub(delta))) - .ok(); -} - -/// A backpressure-aware pipe wrapping tokio's duplex. -/// -/// This provides monitoring and events for backpressure conditions -/// while maintaining compatibility with the standard duplex interface. -pub struct BackpressurePipe { - /// Reader end of the duplex pipe. - reader: DuplexStream, - /// Writer end of the duplex pipe. - writer: DuplexStream, - /// Configuration. - config: ObjectPipeBackpressurePolicy, - /// Current buffer usage (approximate, updated on write). - buffer_usage: AtomicUsize, - /// Current backpressure state. - state: AtomicBool, // true = in high watermark state - /// Total bytes written. - total_written: AtomicUsize, - /// Total bytes read. - total_read: AtomicUsize, - /// Cached high watermark threshold in bytes. - high_watermark_bytes: usize, - /// Cached low watermark threshold in bytes. - low_watermark_bytes: usize, - /// Pipe creation timestamp. - created_at: Instant, -} - -impl BackpressurePipe { - /// Create a new backpressure-aware pipe with default configuration. - pub fn new() -> Self { - Self::with_config(ObjectPipeBackpressurePolicy::from_env()) - } - - /// Create a new backpressure-aware pipe with custom configuration. - pub fn with_config(config: ObjectPipeBackpressurePolicy) -> Self { - let policy = config.to_concurrency_policy(); - let (reader, writer) = duplex(policy.buffer_size); - let high_watermark_bytes = policy.high_watermark_bytes(); - let low_watermark_bytes = policy.low_watermark_bytes(); - - debug!( - buffer_size = config.buffer_size, - high_watermark = config.high_watermark, - low_watermark = config.low_watermark, - high_watermark_bytes, - low_watermark_bytes, - "Created backpressure pipe" - ); - - Self { - reader, - writer, - config, - buffer_usage: AtomicUsize::new(0), - state: AtomicBool::new(false), - total_written: AtomicUsize::new(0), - total_read: AtomicUsize::new(0), - high_watermark_bytes, - low_watermark_bytes, - created_at: Instant::now(), - } - } - - /// Take the reader end of the pipe (consumes self). - pub fn into_reader(self) -> DuplexStream { - self.reader - } - - /// Take the writer end of the pipe (consumes self). - pub fn into_writer(self) -> DuplexStream { - self.writer - } - - /// Split into reader and writer (consumes self). - pub fn split(self) -> (DuplexStream, DuplexStream) { - (self.reader, self.writer) - } - - /// Get current backpressure state. - pub fn state(&self) -> BackpressureState { - if self.state.load(Ordering::Acquire) { - BackpressureState::BackpressureApplied - } else { - BackpressureState::Normal - } - } - - /// Get a compact metadata snapshot for the pipe. - pub fn meta(&self) -> BackpressurePipeMeta { - BackpressurePipeMeta { - buffer_capacity: self.config.buffer_size, - state: self.state(), - age: self.age(), - } - } - - /// Get the age of this pipe. - pub fn age(&self) -> Duration { - self.created_at.elapsed() - } - - /// Get current buffer usage. - pub fn usage(&self) -> usize { - self.buffer_usage.load(Ordering::Acquire) - } - - /// Record bytes written (call after successful write). - pub fn record_write(&self, bytes: usize) { - self.total_written.fetch_add(bytes, Ordering::Relaxed); - self.buffer_usage.fetch_add(bytes, Ordering::Release); - self.update_watermark_state(); - } - - /// Record bytes read (call after successful read). - pub fn record_read(&self, bytes: usize) { - self.total_read.fetch_add(bytes, Ordering::Relaxed); - saturating_sub_atomic(&self.buffer_usage, bytes); - self.update_watermark_state(); - } - - /// Update watermark state and emit transition signals. - fn update_watermark_state(&self) { - let usage = self.buffer_usage.load(Ordering::Acquire); - let usage_percent = calculate_usage_percent(usage, self.config.buffer_size) as u32; - let (next_state, changed) = - apply_watermark_transition(&self.state, usage, self.high_watermark_bytes, self.low_watermark_bytes); - - if changed { - match next_state { - BackpressureState::HighWatermark => { - counter!("rustfs_backpressure_events_total", "state" => "high_watermark").increment(1); - - warn!( - buffer_usage = usage, - buffer_capacity = self.config.buffer_size, - usage_percent, - high_watermark = self.config.high_watermark, - "Backpressure: high watermark reached" - ); - } - BackpressureState::Normal => { - counter!("rustfs_backpressure_events_total", "state" => "normal").increment(1); - - debug!( - buffer_usage = usage, - buffer_capacity = self.config.buffer_size, - usage_percent, - low_watermark = self.config.low_watermark, - "Backpressure: returned to normal" - ); - } - BackpressureState::BackpressureApplied => {} - } - } - } - - /// Get total bytes written. - pub fn total_written(&self) -> usize { - self.total_written.load(Ordering::Relaxed) - } - - /// Get total bytes read. - pub fn total_read(&self) -> usize { - self.total_read.load(Ordering::Relaxed) - } - - /// Get buffer capacity. - pub fn capacity(&self) -> usize { - self.config.buffer_size - } -} - -impl Default for BackpressurePipe { - fn default() -> Self { - Self::new() - } -} - -/// A simple wrapper that provides backpressure monitoring for duplex streams. -/// -/// This is a lighter-weight alternative to `BackpressurePipe` that doesn't -/// wrap the streams but provides monitoring capabilities. -pub struct BackpressureMonitor { - /// Configuration. - config: ObjectPipeBackpressurePolicy, - /// Current buffer usage. - buffer_usage: AtomicUsize, - /// In high watermark state. - in_high_watermark: AtomicBool, - /// Cached high watermark threshold in bytes. - high_watermark_bytes: usize, - /// Cached low watermark threshold in bytes. - low_watermark_bytes: usize, -} - -impl BackpressureMonitor { - /// Create a new monitor with default configuration. - pub fn new() -> Self { - Self::with_config(ObjectPipeBackpressurePolicy::from_env()) - } - - /// Create a new monitor with custom configuration. - pub fn with_config(config: ObjectPipeBackpressurePolicy) -> Self { - let policy = config.to_concurrency_policy(); - let high_watermark_bytes = policy.high_watermark_bytes(); - let low_watermark_bytes = policy.low_watermark_bytes(); - Self { - config, - buffer_usage: AtomicUsize::new(0), - in_high_watermark: AtomicBool::new(false), - high_watermark_bytes, - low_watermark_bytes, - } - } - - /// Record bytes added to buffer. - pub fn on_write(&self, bytes: usize) -> BackpressureState { - self.buffer_usage.fetch_add(bytes, Ordering::Release); - self.update_state() - } - - /// Record bytes removed from buffer. - pub fn on_read(&self, bytes: usize) -> BackpressureState { - saturating_sub_atomic(&self.buffer_usage, bytes); - self.update_state() - } - - /// Get current state. - pub fn state(&self) -> BackpressureState { - if self.in_high_watermark.load(Ordering::Acquire) { - BackpressureState::HighWatermark - } else { - BackpressureState::Normal - } - } - - /// Get current buffer usage. - pub fn usage(&self) -> usize { - self.buffer_usage.load(Ordering::Acquire) - } - - /// Get usage percentage. - pub fn usage_percent(&self) -> f32 { - let usage = self.buffer_usage.load(Ordering::Acquire); - calculate_usage_percent(usage, self.config.buffer_size) - } - - /// Get a compact metadata snapshot for the monitor. - pub fn meta(&self) -> BackpressureMonitorMeta { - let usage = self.buffer_usage.load(Ordering::Acquire); - BackpressureMonitorMeta { - buffer_capacity: self.config.buffer_size, - usage_percent: calculate_usage_percent(usage, self.config.buffer_size), - state: self.state(), - } - } - - /// Update state based on current usage. - fn update_state(&self) -> BackpressureState { - let usage = self.buffer_usage.load(Ordering::Acquire); - let usage_percent = calculate_usage_percent(usage, self.config.buffer_size) as u32; - let (next_state, changed) = - apply_watermark_transition(&self.in_high_watermark, usage, self.high_watermark_bytes, self.low_watermark_bytes); - - if matches!(next_state, BackpressureState::HighWatermark) { - if changed { - counter!("rustfs_backpressure_events_total", "state" => "high_watermark").increment(1); - - debug!(usage_percent, "Backpressure: entered high watermark"); - } - BackpressureState::HighWatermark - } else { - if changed { - counter!("rustfs_backpressure_events_total", "state" => "normal").increment(1); - - debug!(usage_percent, "Backpressure: returned to normal"); - } - BackpressureState::Normal - } - } -} - -impl Default for BackpressureMonitor { - fn default() -> Self { - Self::new() - } -} - -#[cfg(test)] -#[allow(unused_imports)] -mod tests { - use super::{BackpressureMonitor, BackpressurePipe, BackpressureState, ObjectPipeBackpressurePolicy}; - - #[test] - fn test_backpressure_config_default() { - let config = ObjectPipeBackpressurePolicy::default(); - assert_eq!(config.buffer_size, 4 * 1024 * 1024); - assert_eq!(config.high_watermark, 80); - assert_eq!(config.low_watermark, 50); - } - - #[test] - fn test_backpressure_config_watermarks() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 1000, - high_watermark: 80, - low_watermark: 50, - }; - assert_eq!(config.high_watermark_bytes(), 800); - assert_eq!(config.low_watermark_bytes(), 500); - } - - #[test] - fn test_backpressure_policy_projects_to_concurrency_and_core_config() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 2000, - high_watermark: 75, - low_watermark: 40, - }; - let concurrency = config.to_concurrency_policy(); - let core = config.to_core_config(); - - assert_eq!(concurrency.buffer_size, config.buffer_size); - assert_eq!(concurrency.high_watermark, config.high_watermark); - assert_eq!(concurrency.low_watermark, config.low_watermark); - assert_eq!(core.high_water_mark, 0.75); - assert_eq!(core.low_water_mark, 0.40); - assert!(core.enabled); - } - - #[test] - fn test_backpressure_pipe_consumes_concurrency_policy_thresholds() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 2000, - high_watermark: 75, - low_watermark: 40, - }; - let concurrency = config.to_concurrency_policy(); - let pipe = BackpressurePipe::with_config(config); - - assert_eq!(pipe.capacity(), concurrency.buffer_size); - assert_eq!(pipe.high_watermark_bytes, concurrency.high_watermark_bytes()); - assert_eq!(pipe.low_watermark_bytes, concurrency.low_watermark_bytes()); - } - - #[test] - fn test_backpressure_monitor_consumes_concurrency_policy_thresholds() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 2000, - high_watermark: 75, - low_watermark: 40, - }; - let concurrency = config.to_concurrency_policy(); - let monitor = BackpressureMonitor::with_config(config); - - assert_eq!(monitor.meta().buffer_capacity, concurrency.buffer_size); - assert_eq!(monitor.high_watermark_bytes, concurrency.high_watermark_bytes()); - assert_eq!(monitor.low_watermark_bytes, concurrency.low_watermark_bytes()); - } - - #[test] - fn test_backpressure_state_display() { - assert_eq!(format!("{}", BackpressureState::Normal), "normal"); - assert_eq!(format!("{}", BackpressureState::HighWatermark), "high_watermark"); - assert_eq!(format!("{}", BackpressureState::BackpressureApplied), "backpressure_applied"); - } - - #[test] - fn test_backpressure_monitor() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 1000, - high_watermark: 80, - low_watermark: 50, - }; - let monitor = BackpressureMonitor::with_config(config); - - // Initially normal - assert_eq!(monitor.state(), BackpressureState::Normal); - assert_eq!(monitor.meta().buffer_capacity, 1000); - assert_eq!(monitor.meta().usage_percent, 0.0); - - // Write to reach high watermark - let state = monitor.on_write(850); - assert_eq!(state, BackpressureState::HighWatermark); - assert_eq!(monitor.meta().usage_percent, 85.0); - - // Read to go below low watermark - let state = monitor.on_read(400); - assert_eq!(state, BackpressureState::Normal); - assert_eq!(monitor.meta().usage_percent, 45.0); - } - - #[tokio::test] - async fn test_backpressure_pipe_creation() { - let pipe = BackpressurePipe::new(); - assert_eq!(pipe.capacity(), 4 * 1024 * 1024); - assert_eq!(pipe.state(), BackpressureState::Normal); - assert_eq!(pipe.meta().buffer_capacity, 4 * 1024 * 1024); - assert!(pipe.meta().age <= pipe.age()); - } - - #[test] - fn test_backpressure_pipe_state_transitions() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 1000, - high_watermark: 80, - low_watermark: 50, - }; - let pipe = BackpressurePipe::with_config(config); - - assert_eq!(pipe.state(), BackpressureState::Normal); - assert_eq!(pipe.meta().state, BackpressureState::Normal); - - pipe.record_write(850); - assert_eq!(pipe.state(), BackpressureState::BackpressureApplied); - assert_eq!(pipe.meta().state, BackpressureState::BackpressureApplied); - - pipe.record_read(400); - assert_eq!(pipe.state(), BackpressureState::Normal); - assert_eq!(pipe.meta().state, BackpressureState::Normal); - } -} diff --git a/rustfs/src/storage/concurrent_fix_test.rs b/rustfs/src/storage/concurrent_fix_test.rs index 2fedeced1..a80865976 100644 --- a/rustfs/src/storage/concurrent_fix_test.rs +++ b/rustfs/src/storage/concurrent_fix_test.rs @@ -14,17 +14,15 @@ //! Integration tests for concurrent request fix. //! -//! These tests verify that the timeout, backpressure, and deadlock detection -//! mechanisms work correctly under high concurrency scenarios. +//! These tests verify that the timeout and deadlock detection mechanisms work +//! correctly under high concurrency scenarios. #[cfg(test)] mod tests { - use crate::storage::backpressure::{BackpressureMonitor, BackpressureState, ObjectPipeBackpressurePolicy}; use crate::storage::concurrency::{IoLoadLevel, IoPriority}; use crate::storage::deadlock_detector::{ DeadlockDetector, LockInfo, LockType, RequestHangDetectionPolicy, RequestResourceTracker, }; - use crate::storage::lock_optimizer::{LockOptimizeConfig, LockOptimizer, LockStats}; use crate::storage::timeout_wrapper::{GetObjectTimeoutPolicy, RequestTimeoutWrapper, TimedGetObjectResult}; use std::time::Duration; @@ -114,82 +112,6 @@ mod tests { } } - // ============================================ - // Backpressure Tests - // ============================================ - - #[test] - fn test_backpressure_config_defaults() { - let config = ObjectPipeBackpressurePolicy::default(); - assert_eq!(config.buffer_size, 4 * 1024 * 1024); // 4MB - assert_eq!(config.high_watermark, 80); - assert_eq!(config.low_watermark, 50); - } - - #[test] - fn test_backpressure_monitor_state_transitions() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 1000, - high_watermark: 80, - low_watermark: 50, - }; - let monitor = BackpressureMonitor::with_config(config); - - // Initially normal - assert_eq!(monitor.state(), BackpressureState::Normal); - - // Write to reach high watermark - let state = monitor.on_write(850); - assert_eq!(state, BackpressureState::HighWatermark); - - // Read to go below low watermark - let state = monitor.on_read(400); - assert_eq!(state, BackpressureState::Normal); - } - - #[test] - fn test_backpressure_usage_percent() { - let config = ObjectPipeBackpressurePolicy { - buffer_size: 1000, - high_watermark: 80, - low_watermark: 50, - }; - let monitor = BackpressureMonitor::with_config(config); - - monitor.on_write(500); - assert!((monitor.usage_percent() - 50.0).abs() < 1.0); - } - - // ============================================ - // Lock Optimizer Tests - // ============================================ - - #[test] - fn test_lock_optimize_config_defaults() { - let config = LockOptimizeConfig::default(); - assert!(config.enabled); - assert_eq!(config.acquire_timeout, Duration::from_secs(5)); - } - - #[test] - fn test_lock_stats_tracking() { - let stats = LockStats::new(); - - stats.record_acquire(); - stats.record_early_release(Duration::from_millis(100)); - stats.record_early_release(Duration::from_millis(200)); - - assert_eq!(stats.locks_acquired.load(std::sync::atomic::Ordering::Relaxed), 1); - assert_eq!(stats.locks_released_early.load(std::sync::atomic::Ordering::Relaxed), 2); - assert_eq!(stats.max_hold_time(), Duration::from_millis(200)); - } - - #[test] - fn test_lock_optimizer_creation() { - let optimizer = LockOptimizer::new(); - assert!(optimizer.is_enabled()); - } - // ============================================ // I/O Priority Tests // ============================================ diff --git a/rustfs/src/storage/lock_optimizer.rs b/rustfs/src/storage/lock_optimizer.rs deleted file mode 100644 index 3636c50c3..000000000 --- a/rustfs/src/storage/lock_optimizer.rs +++ /dev/null @@ -1,458 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Lock Optimization for GetObject Operations. -//! -//! This module provides optimized lock management for read operations, -//! reducing lock contention by releasing locks early (after metadata read) -//! rather than holding them for the entire data transfer duration. -//! -//! # Migration Note -//! -//! For new code, consider using `rustfs_io_core::LockOptimizer` which provides -//! the same core functionality with better separation of concerns. This module -//! remains for backward compatibility and storage-specific configuration. -//! -//! ```ignore -//! // Recommended: Use io-core directly -//! use rustfs_io_core::LockOptimizer; -//! let optimizer = LockOptimizer::with_defaults(); -//! ``` - -// Allow dead_code for public API that may be used by external modules or future features -//! # Key Features -//! -//! - Early lock release after metadata read -//! - Lock hold time monitoring -//! - Configurable optimization (can be disabled for debugging) -//! - Lock contention metrics emitted through the shared metrics pipeline -//! -//! # Architecture -//! -//! ```text -//! Traditional: [Acquire Lock] --> [Read Metadata] --> [Transfer Data] --> [Release Lock] -//! |<------------------ Lock Held ------------------>| -//! -//! Optimized: [Acquire Lock] --> [Read Metadata] --> [Release Lock] --> [Transfer Data] -//! |<- Lock Held ->| -//! ``` - -use std::sync::Arc; -use std::sync::atomic::{AtomicU64, Ordering}; -use std::time::{Duration, Instant}; -use tracing::debug; - -use metrics::histogram; - -/// Lock optimization configuration. -#[derive(Debug, Clone)] -pub struct LockOptimizeConfig { - /// Whether to enable lock optimization. - /// When enabled, read locks are released after metadata read. - /// When disabled, locks are held for the entire operation (traditional behavior). - pub enabled: bool, - /// Lock acquisition timeout. - pub acquire_timeout: Duration, -} - -impl Default for LockOptimizeConfig { - fn default() -> Self { - Self { - enabled: rustfs_config::DEFAULT_OBJECT_LOCK_OPTIMIZATION_ENABLE, - acquire_timeout: Duration::from_secs(rustfs_config::DEFAULT_OBJECT_LOCK_ACQUIRE_TIMEOUT), - } - } -} - -impl LockOptimizeConfig { - /// Load configuration from environment variables. - pub fn from_env() -> Self { - let enabled = rustfs_utils::get_env_bool( - rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, - rustfs_config::DEFAULT_OBJECT_LOCK_OPTIMIZATION_ENABLE, - ); - let acquire_timeout = Duration::from_secs(rustfs_utils::get_env_u64( - rustfs_config::ENV_OBJECT_LOCK_ACQUIRE_TIMEOUT, - rustfs_config::DEFAULT_OBJECT_LOCK_ACQUIRE_TIMEOUT, - )); - - Self { - enabled, - acquire_timeout, - } - } -} - -/// Statistics for lock optimization monitoring. -#[derive(Debug, Default)] -pub struct LockStats { - /// Total locks acquired. - pub locks_acquired: AtomicU64, - /// Total locks released early. - pub locks_released_early: AtomicU64, - /// Total lock hold time in microseconds. - pub total_hold_time_us: AtomicU64, - /// Maximum lock hold time in microseconds. - pub max_hold_time_us: AtomicU64, -} - -impl LockStats { - /// Create new lock statistics. - pub fn new() -> Self { - Self::default() - } - - /// Record a lock acquisition. - pub fn record_acquire(&self) { - self.locks_acquired.fetch_add(1, Ordering::Relaxed); - } - - /// Record an early lock release. - pub fn record_early_release(&self, hold_time: Duration) { - self.locks_released_early.fetch_add(1, Ordering::Relaxed); - self.record_hold_time(hold_time); - } - - /// Record lock hold time. - fn record_hold_time(&self, hold_time: Duration) { - let hold_time_us = hold_time.as_micros() as u64; - self.total_hold_time_us.fetch_add(hold_time_us, Ordering::Relaxed); - - // Update max hold time - let mut current_max = self.max_hold_time_us.load(Ordering::Relaxed); - while hold_time_us > current_max { - match self - .max_hold_time_us - .compare_exchange_weak(current_max, hold_time_us, Ordering::Relaxed, Ordering::Relaxed) - { - Ok(_) => break, - Err(actual) => current_max = actual, - } - } - } - - /// Get average hold time. - pub fn avg_hold_time(&self) -> Duration { - let total = self.total_hold_time_us.load(Ordering::Relaxed); - let count = self.locks_released_early.load(Ordering::Relaxed); - total.checked_div(count).map(Duration::from_micros).unwrap_or(Duration::ZERO) - } - - /// Get maximum hold time. - pub fn max_hold_time(&self) -> Duration { - Duration::from_micros(self.max_hold_time_us.load(Ordering::Relaxed)) - } -} - -/// Global lock statistics. -static LOCK_STATS: std::sync::OnceLock> = std::sync::OnceLock::new(); - -/// Get global lock statistics. -pub fn get_lock_stats() -> Arc { - LOCK_STATS.get_or_init(|| Arc::new(LockStats::new())).clone() -} - -/// An optimized lock guard that supports early release. -/// -/// This wraps the actual lock guard and provides: -/// - Early release capability (before drop) -/// - Hold time tracking -/// - Metrics reporting -pub struct OptimizedLockGuard { - /// The underlying lock guard. - guard: Option, - /// When the lock was acquired. - acquire_time: Instant, - /// Whether the lock has been released. - released: bool, - /// Lock resource name (for logging). - resource: String, - /// Statistics reference. - stats: Arc, -} - -impl OptimizedLockGuard { - /// Create a new optimized lock guard. - pub fn new(guard: G, resource: impl Into) -> Self { - let stats = get_lock_stats(); - stats.record_acquire(); - - Self { - guard: Some(guard), - acquire_time: Instant::now(), - released: false, - resource: resource.into(), - stats, - } - } - - /// Get the lock hold time so far. - pub fn hold_time(&self) -> Duration { - self.acquire_time.elapsed() - } - - /// Check if the lock has been released. - pub fn is_released(&self) -> bool { - self.released - } - - /// Release the lock early (before drop). - /// - /// This is the key optimization: releasing the lock after - /// metadata read rather than waiting for the entire operation. - pub fn early_release(&mut self) { - if self.released { - return; - } - - let hold_time = self.hold_time(); - self.guard.take(); - self.released = true; - - self.stats.record_early_release(hold_time); - - histogram!("rustfs_lock_hold_duration_seconds").record(hold_time.as_secs_f64()); - - debug!( - resource = %self.resource, - hold_time_ms = hold_time.as_millis(), - "Lock released early (optimization active)" - ); - } - - /// Get a reference to the underlying guard. - pub fn as_ref(&self) -> Option<&G> { - if self.released { None } else { self.guard.as_ref() } - } -} - -impl Drop for OptimizedLockGuard { - fn drop(&mut self) { - if !self.released { - let hold_time = self.hold_time(); - self.guard.take(); - self.released = true; - - self.stats.record_early_release(hold_time); - - histogram!("rustfs_lock_hold_duration_seconds").record(hold_time.as_secs_f64()); - - debug!( - resource = %self.resource, - hold_time_ms = hold_time.as_millis(), - "Lock released on drop (normal release)" - ); - } - } -} - -/// A scope guard that releases a lock when it goes out of scope. -/// -/// This is a simpler version of OptimizedLockGuard for cases -/// where we just need RAII semantics without tracking. -pub struct LockScopeGuard { - guard: Option, -} - -impl LockScopeGuard { - /// Create a new scope guard. - pub fn new(guard: G) -> Self { - Self { guard: Some(guard) } - } - - /// Release the lock early. - pub fn release(&mut self) { - self.guard.take(); - } -} - -impl Drop for LockScopeGuard { - fn drop(&mut self) { - self.guard.take(); - } -} - -/// Helper for managing lock optimization in GetObject operations. -/// -/// This provides a clean interface for the common pattern: -/// 1. Acquire lock -/// 2. Read metadata -/// 3. Release lock (if optimization enabled) -/// 4. Transfer data (without lock) -pub struct LockOptimizer { - /// Configuration. - config: LockOptimizeConfig, -} - -impl LockOptimizer { - /// Create a new lock optimizer with default configuration. - pub fn new() -> Self { - Self { - config: LockOptimizeConfig::from_env(), - } - } - - /// Create a new lock optimizer with custom configuration. - pub fn with_config(config: LockOptimizeConfig) -> Self { - Self { config } - } - - /// Check if lock optimization is enabled. - pub fn is_enabled(&self) -> bool { - self.config.enabled - } - - /// Get the lock acquisition timeout. - pub fn acquire_timeout(&self) -> Duration { - self.config.acquire_timeout - } - - /// Wrap a lock guard for optimization. - pub fn wrap_guard(&self, guard: G, resource: impl Into) -> OptimizedLockGuard { - OptimizedLockGuard::new(guard, resource) - } - - /// Execute a metadata read operation with lock optimization. - /// - /// This is the main entry point for optimized lock usage: - /// - If optimization is enabled: lock is released after metadata_fn completes - /// - If optimization is disabled: lock is held until the returned guard is dropped - /// - /// # Arguments - /// - /// * `guard` - The lock guard to optimize - /// * `resource` - Resource name for logging - /// * `metadata_fn` - Function to read metadata while holding lock - /// - /// # Returns - /// - /// A tuple of (metadata result, optional guard to hold for later release) - pub async fn with_optimized_lock( - &self, - guard: G, - resource: impl Into, - metadata_fn: F, - ) -> (T, Option>) - where - F: FnOnce() -> Fut, - Fut: std::future::Future, - { - let resource = resource.into(); - let mut optimized = OptimizedLockGuard::new(guard, &resource); - - // Execute metadata read while holding lock - let result = metadata_fn().await; - - if self.config.enabled { - // Release lock early - optimized.early_release(); - (result, None) - } else { - // Keep lock for caller to release - (result, Some(optimized)) - } - } -} - -impl Default for LockOptimizer { - fn default() -> Self { - Self::new() - } -} - -/// Check if lock optimization is enabled globally. -pub fn is_lock_optimization_enabled() -> bool { - rustfs_utils::get_env_bool( - rustfs_config::ENV_OBJECT_LOCK_OPTIMIZATION_ENABLE, - rustfs_config::DEFAULT_OBJECT_LOCK_OPTIMIZATION_ENABLE, - ) -} - -#[cfg(test)] -#[allow(unused_imports)] -mod tests { - use super::{LockOptimizeConfig, LockOptimizer, LockStats, OptimizedLockGuard}; - use std::sync::Mutex; - use std::sync::atomic::Ordering; - use std::time::Duration; - - #[test] - fn test_lock_optimize_config_default() { - let config = LockOptimizeConfig::default(); - assert!(config.enabled); - assert_eq!(config.acquire_timeout, Duration::from_secs(5)); - } - - #[test] - fn test_lock_stats() { - let stats = LockStats::new(); - - stats.record_acquire(); - stats.record_early_release(Duration::from_millis(100)); - stats.record_early_release(Duration::from_millis(200)); - - assert_eq!(stats.locks_acquired.load(Ordering::Relaxed), 1); - assert_eq!(stats.locks_released_early.load(Ordering::Relaxed), 2); - assert_eq!(stats.max_hold_time(), Duration::from_millis(200)); - } - - #[test] - fn test_optimized_lock_guard() { - let guard = Mutex::new(42); - let locked = guard.lock().unwrap(); - let mut optimized = OptimizedLockGuard::new(locked, "test-resource"); - - assert!(!optimized.is_released()); - assert!(optimized.hold_time() < Duration::from_secs(1)); - - optimized.early_release(); - assert!(optimized.is_released()); - } - - #[test] - fn test_lock_optimizer() { - let optimizer = LockOptimizer::new(); - assert!(optimizer.is_enabled()); - } - - #[tokio::test] - async fn test_with_optimized_lock_enabled() { - let optimizer = LockOptimizer::new(); - let guard = Mutex::new(42); - let locked = guard.lock().unwrap(); - - let (result, returned_guard) = optimizer.with_optimized_lock(locked, "test-resource", || async { 100 }).await; - - assert_eq!(result, 100); - // With optimization enabled, guard should be None (released early) - assert!(returned_guard.is_none()); - } - - #[tokio::test] - async fn test_with_optimized_lock_disabled() { - let config = LockOptimizeConfig { - enabled: false, - acquire_timeout: Duration::from_secs(5), - }; - let optimizer = LockOptimizer::with_config(config); - let guard = Mutex::new(42); - let locked = guard.lock().unwrap(); - - let (result, returned_guard) = optimizer.with_optimized_lock(locked, "test-resource", || async { 100 }).await; - - assert_eq!(result, 100); - // With optimization disabled, guard should be Some (held for later) - assert!(returned_guard.is_some()); - } -} diff --git a/rustfs/src/storage/mod.rs b/rustfs/src/storage/mod.rs index 58b6a0ce9..e0817da90 100644 --- a/rustfs/src/storage/mod.rs +++ b/rustfs/src/storage/mod.rs @@ -13,12 +13,10 @@ // limitations under the License. pub mod access; -pub mod backpressure; pub mod concurrency; pub mod deadlock_detector; pub mod ecfs; pub(crate) mod helper; -pub mod lock_optimizer; pub mod options; pub mod request_context; pub mod rpc; From 51497cb5335b5f02bdf16355bb413bad182da757 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 12:57:44 +0800 Subject: [PATCH 02/24] fix(ecstore): give peer REST failures op and bucket context (#6200) --- .../src/cluster/rpc/peer_rest_client.rs | 196 +++++++++++++++--- .../ecstore/src/cluster/rpc/peer_s3_client.rs | 2 + 2 files changed, 166 insertions(+), 32 deletions(-) diff --git a/crates/ecstore/src/cluster/rpc/peer_rest_client.rs b/crates/ecstore/src/cluster/rpc/peer_rest_client.rs index 1426fc91f..3bac43266 100644 --- a/crates/ecstore/src/cluster/rpc/peer_rest_client.rs +++ b/crates/ecstore/src/cluster/rpc/peer_rest_client.rs @@ -86,6 +86,25 @@ const PEER_REST_RECOVERY_MAX_BACKOFF: Duration = Duration::from_secs(30); const SCANNER_ACTIVITY_MAX_MESSAGE_SIZE: usize = 1024; const REPLICATION_STATS_MAX_MESSAGE_SIZE: usize = 8 * 1024 * 1024; +/// Error for a peer that reported `success = false` without an `error_info` payload. +/// +/// Same shape as `peer_s3_client::peer_failure_without_details`, over `StorageError` +/// instead of `DiskError`. The message names the operation (and the bucket, where the +/// operation has one) and nothing else, for two reasons: +/// +/// - `finalize_result` classifies failures by message substring, so any text matching +/// `message_has_network_needle` would take an answering peer offline and evict its +/// connection over a plain application-level rejection. +/// - Quorum aggregation (`reduce_errs`) buckets `Io` errors by kind plus rendered +/// message, so a per-peer detail such as the peer address would split one shared +/// failure into single-count buckets and downgrade the dominant error. +fn peer_failure_without_details(op: &str, bucket: Option<&str>) -> Error { + match bucket { + Some(bucket) => Error::other(format!("{op}({bucket}): peer returned failure without error details")), + None => Error::other(format!("{op}: peer returned failure without error details")), + } +} + fn decode_bucket_stats_response(response: GetBucketStatsDataResponse) -> Result { if !response.success { return Err(Error::other( @@ -696,7 +715,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("local_storage_info", None)); } let data = response.storage_info; @@ -719,7 +738,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("server_info", None)); } let data = response.server_properties; @@ -742,7 +761,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_cpus", None)); } let data = response.cpus; @@ -765,7 +784,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_net_info", None)); } let data = response.net_info; @@ -788,7 +807,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_partitions", None)); } let data = response.partitions; @@ -811,7 +830,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_os_info", None)); } let data = response.os_info; @@ -832,7 +851,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_se_linux_info", None)); } let data = response.sys_services; @@ -857,7 +876,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_sys_config", None)); } let data = response.sys_config; @@ -882,7 +901,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_sys_errors", None)); } let data = response.sys_errors; @@ -907,7 +926,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_mem_info", None)); } let data = response.mem_info; @@ -939,7 +958,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_metrics", None)); } let data = response.realtime_metrics; @@ -964,7 +983,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_live_events", None)); } Ok(PeerLiveEventsBatch { @@ -989,7 +1008,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("get_proc_info", None)); } let data = response.proc_info; @@ -1016,7 +1035,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("start_profiling", None)); } Ok(()) } @@ -1323,7 +1342,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket))); } Ok(()) } @@ -1346,7 +1365,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("delete_bucket_metadata", Some(bucket))); } Ok(()) } @@ -1369,7 +1388,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("delete_policy", None)); } Ok(()) } @@ -1392,7 +1411,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_policy", None)); } Ok(()) } @@ -1417,7 +1436,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_policy_mapping", None)); } Ok(()) } @@ -1440,7 +1459,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("delete_user", None)); } Ok(()) } @@ -1463,7 +1482,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("delete_service_account", None)); } Ok(()) } @@ -1487,7 +1506,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_user", None)); } Ok(()) } @@ -1510,7 +1529,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_service_account", None)); } Ok(()) } @@ -1533,7 +1552,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_group", None)); } Ok(()) } @@ -1554,7 +1573,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("reload_site_replication_config", None)); } Ok(()) } @@ -1597,7 +1616,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("signal_service", None)); } validate_signal_service_protocol(sig, sub_sys, response.protocol_version)?; Ok(response) @@ -1667,7 +1686,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("reload_pool_meta", None)); } Ok(()) @@ -1691,7 +1710,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("stop_rebalance", None)); } Ok(()) @@ -1725,7 +1744,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("load_rebalance_meta", None)); } Ok(()) @@ -1753,7 +1772,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("start_decommission", None)); } Ok(()) @@ -1777,7 +1796,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("decommission_cancel", None)); } Ok(()) @@ -1801,7 +1820,7 @@ impl PeerRestClient { if let Some(msg) = response.error_info { return Err(Error::other(msg)); } - return Err(Error::other("")); + return Err(peer_failure_without_details("clear_decommission", None)); } Ok(()) @@ -1947,6 +1966,8 @@ fn tier_config_reload_status_outcome(status: tonic::Status) -> TierConfigReloadO mod tests { use super::*; use crate::config::com::STORAGE_CLASS_SUB_SYS; + use crate::disk::error::DiskError; + use crate::disk::error_reduce::reduce_errs; use crate::layout::{disks_layout::DisksLayout, endpoints::SetupType}; use rustfs_config::{ENV_KUBERNETES_SERVICE_HOST, ENV_LOCAL_ENDPOINT_HOST, ENV_STARTUP_TOPOLOGY_WAIT_MODE}; use serde_json::Value; @@ -3098,4 +3119,115 @@ mod tests { && span.get("request_id").and_then(Value::as_str) == Some("req-peer-rest") })); } + + /// Every operation name passed to `peer_failure_without_details` in this file. + const PEER_FAILURE_OPS: &[&str] = &[ + "local_storage_info", + "server_info", + "get_cpus", + "get_net_info", + "get_partitions", + "get_os_info", + "get_se_linux_info", + "get_sys_config", + "get_sys_errors", + "get_mem_info", + "get_metrics", + "get_live_events", + "get_proc_info", + "start_profiling", + "load_bucket_metadata", + "delete_bucket_metadata", + "delete_policy", + "load_policy", + "load_policy_mapping", + "delete_user", + "delete_service_account", + "load_user", + "load_service_account", + "load_group", + "reload_site_replication_config", + "signal_service", + "reload_pool_meta", + "stop_rebalance", + "load_rebalance_meta", + "start_decommission", + "decommission_cancel", + "clear_decommission", + ]; + + #[test] + fn peer_failure_without_details_names_operation_and_bucket() { + for op in PEER_FAILURE_OPS { + let message = peer_failure_without_details(op, None).to_string(); + assert!(message.contains(op), "{op} message must name the operation: {message}"); + } + + for op in ["load_bucket_metadata", "delete_bucket_metadata"] { + let message = peer_failure_without_details(op, Some("ops-bucket")).to_string(); + assert!(message.contains(op), "{op} message must name the operation: {message}"); + assert!(message.contains("ops-bucket"), "{op} message must name the bucket: {message}"); + } + } + + #[test] + fn peer_failure_without_details_keeps_one_reduce_errs_bucket_per_operation() { + // reduce_errs groups Io errors by kind plus rendered message: peers failing the + // same operation must stay a single dominant error instead of one bucket per peer. + let per_peer_errs = (0..4) + .map(|_| Some(DiskError::from(peer_failure_without_details("load_bucket_metadata", Some("shared"))))) + .collect::>(); + let (count, dominant) = reduce_errs(&per_peer_errs, &[]); + assert_eq!(count, 4, "one shared failure must not split into per-peer buckets"); + assert_eq!( + dominant, + Some(DiskError::from(peer_failure_without_details("load_bucket_metadata", Some("shared")))) + ); + + assert_ne!( + peer_failure_without_details("load_bucket_metadata", Some("shared")).to_string(), + peer_failure_without_details("delete_bucket_metadata", Some("shared")).to_string() + ); + assert_ne!( + peer_failure_without_details("load_bucket_metadata", Some("bucket-a")).to_string(), + peer_failure_without_details("load_bucket_metadata", Some("bucket-b")).to_string() + ); + } + + #[test] + fn peer_failure_without_details_never_reads_as_a_network_failure() { + // `finalize_result` marks the peer offline and evicts its connection whenever the + // message matches a network needle. A peer that answered `success = false` is alive, + // so no operation or bucket name may push this text over that classifier. + for op in PEER_FAILURE_OPS { + let err = peer_failure_without_details(op, None); + assert!( + !PeerRestClient::is_network_like_error(&err), + "{op} must not read as a transport failure: {err}" + ); + + let scoped = peer_failure_without_details(op, Some("bucket-name")); + assert!( + !PeerRestClient::is_network_like_error(&scoped), + "{op} must not read as a transport failure: {scoped}" + ); + } + + // The bucket name is caller-supplied. Every needle carries a space, which S3 bucket + // names cannot, and the name is closed by `)` before the literal text resumes, so no + // needle can straddle the boundary either. + for bucket in [ + "timed-out", + "connection-reset", + "transport-error", + "broken-pipe", + "unavailable-logs", + ] { + let err = peer_failure_without_details("load_bucket_metadata", Some(bucket)); + assert!( + !PeerRestClient::is_network_like_error(&err), + "bucket {bucket} must not push the message over the network classifier: {err}" + ); + } + } } diff --git a/crates/ecstore/src/cluster/rpc/peer_s3_client.rs b/crates/ecstore/src/cluster/rpc/peer_s3_client.rs index 02c9f75f1..d44651dce 100644 --- a/crates/ecstore/src/cluster/rpc/peer_s3_client.rs +++ b/crates/ecstore/src/cluster/rpc/peer_s3_client.rs @@ -220,6 +220,8 @@ fn pool_write_quorum(participant_count: usize) -> usize { /// buckets `Error::Io` by kind plus rendered message, so any per-peer detail (address, /// timing) would split one shared failure into single-count buckets and downgrade a real /// dominant error into `ErasureWriteQuorum`. +/// +/// `peer_rest_client` carries the same helper over `StorageError` for the same response shape. fn peer_failure_without_details(op: &str, bucket: Option<&str>) -> Error { match bucket { Some(bucket) => Error::other(format!("{op}({bucket}): peer returned failure without error details")), From 9ef059c9083f637e45666ec95ad6263fdaa5723b Mon Sep 17 00:00:00 2001 From: hector <42570491+majinghe@users.noreply.github.com> Date: Tue, 18 Aug 2026 13:18:13 +0800 Subject: [PATCH 03/24] ci(package): auto-trigger DEB/RPM packaging on releases and upload to GitHub release assets (#6202) --- .github/workflows/package.yml | 95 +++++++++++++++++++++++++++++++---- 1 file changed, 84 insertions(+), 11 deletions(-) diff --git a/.github/workflows/package.yml b/.github/workflows/package.yml index d44205fa0..b9536d263 100644 --- a/.github/workflows/package.yml +++ b/.github/workflows/package.yml @@ -15,28 +15,35 @@ # Package Workflow - Build DEB/RPM packages # # This workflow builds DEB and RPM packages from pre-built Linux binaries -# and uploads them to Cloudflare R2. +# and uploads them to Cloudflare R2 and the GitHub release. # # Trigger: -# - release published: automatically package when a GitHub release is published -# - workflow_dispatch: manual trigger with optional tag/run_id +# - workflow_run: automatically package after "Build and Release" completes +# for a release tag (the mac/windows/linux binaries are already uploaded +# to the GitHub release before packaging starts) +# - workflow_dispatch: manual fallback (backfill / re-run) with optional tag/run_id # # Flow: -# 1. Find the Build workflow run for the release tag +# 1. Resolve the triggering Build workflow run for the release tag # 2. Download Linux binaries (x86_64-gnu, aarch64-gnu) from build artifacts # 3. Build DEB packages for amd64 and arm64 # 4. Build RPM packages for x86_64 and aarch64 -# 5. Upload all packages to Cloudflare R2 +# 5. Upload all packages to Cloudflare R2 and the GitHub release name: Package DEB/RPM permissions: - contents: read + # contents: write is required to upload packages to the GitHub release + contents: write actions: read on: - release: - types: [ published ] + # Follows the same pattern as docker.yml: run after the release build + # workflow completes, so packaging is triggered only by release tags + # (e.g. 1.0.0-rc.2, 1.0.0-rc.3), never by development builds. + workflow_run: + workflows: [ "Build and Release" ] + types: [ completed ] workflow_dispatch: inputs: tag: @@ -49,13 +56,26 @@ on: type: string concurrency: - group: ${{ github.workflow }}-${{ github.event.release.tag_name || github.event.inputs.tag || github.run_id }} + group: ${{ github.workflow }}-${{ github.event.workflow_run.head_branch || github.event.inputs.tag || github.run_id }} cancel-in-progress: true +env: + HEAD_BRANCH: ${{ github.event.workflow_run.head_branch }} + WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }} + jobs: # Resolve which build run to use and extract version info resolve: name: Resolve Build + # Auto-trigger only from successful tag builds of "Build and Release". + # Tag pushes arrive as event == push with head_branch != main (a + # non-main push head_branch is the release tag name). Manual dispatch + # stays available as a fallback for backfills and re-runs. + if: >- + github.event_name == 'workflow_dispatch' || + (github.event.workflow_run.conclusion == 'success' && + github.event.workflow_run.event == 'push' && + github.event.workflow_run.head_branch != 'main') runs-on: ubuntu-latest timeout-minutes: 10 outputs: @@ -75,8 +95,8 @@ jobs: set -euo pipefail # Determine tag - if [[ "${{ github.event_name }}" == "release" ]]; then - TAG="${{ github.event.release.tag_name }}" + if [[ "${{ github.event_name }}" == "workflow_run" ]]; then + TAG="${HEAD_BRANCH}" elif [[ -n "$INPUT_TAG" ]]; then TAG="$INPUT_TAG" else @@ -93,6 +113,11 @@ jobs: BUILD_RUN_ID="$INPUT_RUN_ID" echo "Using explicit build run ID: $BUILD_RUN_ID" + elif [[ "${{ github.event_name }}" == "workflow_run" ]]; then + # Use the Build and Release run that triggered this workflow + BUILD_RUN_ID="${WORKFLOW_RUN_ID}" + echo "Using triggering workflow run: $BUILD_RUN_ID" + elif [[ -n "$TAG" ]]; then # Find the build run that produced this tag echo "Looking for build run for tag: $TAG" @@ -456,6 +481,54 @@ jobs: echo "✅ Latest packages updated" fi + - name: Upload packages to GitHub Release + if: needs.resolve.outputs.tag != '' + env: + GH_TOKEN: ${{ github.token }} + shell: bash + run: | + set -euo pipefail + + TAG="${{ needs.resolve.outputs.tag }}" + DEB_FILE="${{ steps.deb.outputs.deb_file }}" + RPM_FILE="${{ steps.rpm.outputs.rpm_file }}" + + # Upload the packages, then refresh the release checksums so the new + # assets are covered, matching the binary release flow. + for f in "$DEB_FILE" "$RPM_FILE"; do + if [[ -n "$f" && -f "$f" ]]; then + echo "📤 Uploading $(basename "$f") to GitHub release ${TAG}..." + gh release upload "$TAG" "$f" --clobber + fi + done + + CHECKSUM_DIR="$(mktemp -d)" + gh release download "$TAG" -p 'SHA256SUMS' -p 'SHA512SUMS' \ + -D "$CHECKSUM_DIR" --clobber 2>/dev/null || true + + for spec in "SHA256SUMS:sha256sum" "SHA512SUMS:sha512sum"; do + asset="${spec%%:*}" + checksum_cmd="${spec##*:}" + checksum_file="${CHECKSUM_DIR}/${asset}" + + touch "$checksum_file" + + for f in "$DEB_FILE" "$RPM_FILE"; do + if [[ -n "$f" && -f "$f" ]]; then + base="$(basename "$f")" + # Remove any stale entry, then append the fresh digest + grep -Fv -- "$base" "$checksum_file" > "${checksum_file}.tmp" || true + mv "${checksum_file}.tmp" "$checksum_file" + (cd "$(dirname "$f")" && "$checksum_cmd" -- "$base") >> "$checksum_file" + fi + done + + echo "📤 Updating ${asset} for release ${TAG}..." + gh release upload "$TAG" "$checksum_file" --clobber + done + + echo "✅ GitHub release assets updated" + # Summary summary: name: Summary From 60eb139db9306f6ada4ed5090437a86bd41d5e95 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 13:27:38 +0800 Subject: [PATCH 04/24] refactor: import x-amz-checksum header names from the shared constants (#6193) Co-authored-by: houseme --- crates/ecstore/src/client/api_get_options.rs | 3 +- crates/ecstore/src/client/transition_api.rs | 16 ++++++---- rustfs/src/admin/router.rs | 31 ++++++++++---------- 3 files changed, 27 insertions(+), 23 deletions(-) diff --git a/crates/ecstore/src/client/api_get_options.rs b/crates/ecstore/src/client/api_get_options.rs index 503b44f3a..3438453b3 100644 --- a/crates/ecstore/src/client/api_get_options.rs +++ b/crates/ecstore/src/client/api_get_options.rs @@ -20,6 +20,7 @@ #![allow(clippy::all)] use http::{HeaderMap, HeaderName, HeaderValue}; +use rustfs_utils::http::headers::AMZ_CHECKSUM_MODE; use std::collections::HashMap; use time::OffsetDateTime; use tracing::warn; @@ -76,7 +77,7 @@ impl GetObjectOptions { } } if self.checksum { - headers.insert(HeaderName::from_static("x-amz-checksum-mode"), HeaderValue::from_static("ENABLED")); + headers.insert(HeaderName::from_static(AMZ_CHECKSUM_MODE), HeaderValue::from_static("ENABLED")); } headers } diff --git a/crates/ecstore/src/client/transition_api.rs b/crates/ecstore/src/client/transition_api.rs index 6ad802418..be8e0a510 100644 --- a/crates/ecstore/src/client/transition_api.rs +++ b/crates/ecstore/src/client/transition_api.rs @@ -54,6 +54,10 @@ use rustfs_config::MAX_S3_CLIENT_RESPONSE_SIZE; use rustfs_rio::HashReader; use rustfs_utils::HashAlgorithm; use rustfs_utils::{ + http::headers::{ + AMZ_CHECKSUM_CRC32, AMZ_CHECKSUM_CRC32C, AMZ_CHECKSUM_CRC64NVME, AMZ_CHECKSUM_MODE, AMZ_CHECKSUM_SHA1, + AMZ_CHECKSUM_SHA256, + }, net::get_endpoint_url, retry::{DEFAULT_RETRY_CAP, DEFAULT_RETRY_UNIT, MAX_JITTER, MAX_RETRY, RetryTimer}, }; @@ -1383,12 +1387,12 @@ pub(crate) fn to_object_info_for_provider( }; // Extract checksums - let checksum_crc32 = get_header("x-amz-checksum-crc32"); - let checksum_crc32c = get_header("x-amz-checksum-crc32c"); - let checksum_sha1 = get_header("x-amz-checksum-sha1"); - let checksum_sha256 = get_header("x-amz-checksum-sha256"); - let checksum_crc64nvme = get_header("x-amz-checksum-crc64nvme"); - let checksum_mode = get_header("x-amz-checksum-mode"); + let checksum_crc32 = get_header(AMZ_CHECKSUM_CRC32); + let checksum_crc32c = get_header(AMZ_CHECKSUM_CRC32C); + let checksum_sha1 = get_header(AMZ_CHECKSUM_SHA1); + let checksum_sha256 = get_header(AMZ_CHECKSUM_SHA256); + let checksum_crc64nvme = get_header(AMZ_CHECKSUM_CRC64NVME); + let checksum_mode = get_header(AMZ_CHECKSUM_MODE); // Build and return the ObjectInfo struct Ok(ObjectInfo { diff --git a/rustfs/src/admin/router.rs b/rustfs/src/admin/router.rs index 9eb893792..717de0bee 100644 --- a/rustfs/src/admin/router.rs +++ b/rustfs/src/admin/router.rs @@ -67,6 +67,9 @@ use rustfs_policy::policy::action::{Action, S3Action}; use rustfs_s3_types::EventName; use rustfs_signer::pre_sign_v4; use rustfs_utils::egress::{OutboundDnsResolver, OutboundPolicy}; +use rustfs_utils::http::headers::{ + AMZ_CHECKSUM_CRC32, AMZ_CHECKSUM_CRC32C, AMZ_CHECKSUM_CRC64NVME, AMZ_CHECKSUM_SHA1, AMZ_CHECKSUM_SHA256, AMZ_CHECKSUM_TYPE, +}; use rustfs_utils::http::{ SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_REPLICATION_CHECK, SUFFIX_SOURCE_REPLICATION_REQUEST, SUFFIX_SOURCE_VERSION_ID, get_source_scheme, insert_header, @@ -1031,28 +1034,24 @@ fn build_get_object_response_headers(output: &GetObjectOutput, base_headers: &He )?; } if let Some(checksum_crc32) = &output.checksum_crc32 { - insert_string_header(&mut headers, HeaderName::from_static("x-amz-checksum-crc32"), checksum_crc32.clone())?; + insert_string_header(&mut headers, HeaderName::from_static(AMZ_CHECKSUM_CRC32), checksum_crc32.clone())?; } if let Some(checksum_crc32c) = &output.checksum_crc32c { - insert_string_header(&mut headers, HeaderName::from_static("x-amz-checksum-crc32c"), checksum_crc32c.clone())?; + insert_string_header(&mut headers, HeaderName::from_static(AMZ_CHECKSUM_CRC32C), checksum_crc32c.clone())?; } if let Some(checksum_crc64nvme) = &output.checksum_crc64nvme { - insert_string_header( - &mut headers, - HeaderName::from_static("x-amz-checksum-crc64nvme"), - checksum_crc64nvme.clone(), - )?; + insert_string_header(&mut headers, HeaderName::from_static(AMZ_CHECKSUM_CRC64NVME), checksum_crc64nvme.clone())?; } if let Some(checksum_sha1) = &output.checksum_sha1 { - insert_string_header(&mut headers, HeaderName::from_static("x-amz-checksum-sha1"), checksum_sha1.clone())?; + insert_string_header(&mut headers, HeaderName::from_static(AMZ_CHECKSUM_SHA1), checksum_sha1.clone())?; } if let Some(checksum_sha256) = &output.checksum_sha256 { - insert_string_header(&mut headers, HeaderName::from_static("x-amz-checksum-sha256"), checksum_sha256.clone())?; + insert_string_header(&mut headers, HeaderName::from_static(AMZ_CHECKSUM_SHA256), checksum_sha256.clone())?; } if let Some(checksum_type) = &output.checksum_type { insert_string_header( &mut headers, - HeaderName::from_static("x-amz-checksum-type"), + HeaderName::from_static(AMZ_CHECKSUM_TYPE), checksum_type.as_str().to_string(), )?; } @@ -1114,12 +1113,12 @@ fn clear_object_lambda_variant_headers(headers: &mut HeaderMap) { http::header::ETAG, http::header::LAST_MODIFIED, http::header::EXPIRES, - HeaderName::from_static("x-amz-checksum-crc32"), - HeaderName::from_static("x-amz-checksum-crc32c"), - HeaderName::from_static("x-amz-checksum-crc64nvme"), - HeaderName::from_static("x-amz-checksum-sha1"), - HeaderName::from_static("x-amz-checksum-sha256"), - HeaderName::from_static("x-amz-checksum-type"), + HeaderName::from_static(AMZ_CHECKSUM_CRC32), + HeaderName::from_static(AMZ_CHECKSUM_CRC32C), + HeaderName::from_static(AMZ_CHECKSUM_CRC64NVME), + HeaderName::from_static(AMZ_CHECKSUM_SHA1), + HeaderName::from_static(AMZ_CHECKSUM_SHA256), + HeaderName::from_static(AMZ_CHECKSUM_TYPE), HeaderName::from_static("x-amz-tagging-count"), HeaderName::from_static("x-amz-request-route"), HeaderName::from_static("x-amz-request-token"), From 355c8d2e2292596bbbb034fca1fbcec998dc935a Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 13:40:10 +0800 Subject: [PATCH 05/24] fix(admin): classify missing kms config by error variant (#6196) --- .../ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs | 2 +- crates/ecstore/src/error/mod.rs | 8 ++++++++ rustfs/src/admin/handlers/kms_dynamic.rs | 8 ++++++-- 3 files changed, 15 insertions(+), 3 deletions(-) diff --git a/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs b/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs index ec5aaf74e..326b798ee 100644 --- a/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs +++ b/crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs @@ -1783,7 +1783,7 @@ impl TransitionState { .await; } global_metrics().record_scanner_transition_failed(1); - if !is_err_version_not_found(&err) && !is_err_object_not_found(&err) && !is_network_or_host_down(&err.to_string(), false) && !err.to_string().contains("use of closed network connection") { + if !is_err_version_not_found(&err) && !is_err_object_not_found(&err) && !is_network_or_host_down(&err.to_string(), false) { error!( event = EVENT_LIFECYCLE_TIER_OPERATION_FAILED, component = LOG_COMPONENT_ECSTORE, diff --git a/crates/ecstore/src/error/mod.rs b/crates/ecstore/src/error/mod.rs index 76dcc7a0c..f8215d908 100644 --- a/crates/ecstore/src/error/mod.rs +++ b/crates/ecstore/src/error/mod.rs @@ -1116,6 +1116,14 @@ mod tests { assert!(encoder_source.is::()); } + // The lifecycle transition worker relies on this arm alone to suppress the + // closed-connection noise (`bucket_lifecycle_ops.rs`); dropping it here would + // silently turn shutdown races back into `error!` log spam. + #[test] + fn is_network_or_host_down_covers_closed_network_connection() { + assert!(is_network_or_host_down("transition failed: use of closed network connection", false)); + } + // Regression for #952 (ECA-11): an all-`DiskNotFound` slice (every drive in // every set unreachable) must NOT be classified as "all not found", // otherwise ListObjects silently returns an empty listing and masks a full diff --git a/rustfs/src/admin/handlers/kms_dynamic.rs b/rustfs/src/admin/handlers/kms_dynamic.rs index 887ec7fad..d7ee7a0f7 100644 --- a/rustfs/src/admin/handlers/kms_dynamic.rs +++ b/rustfs/src/admin/handlers/kms_dynamic.rs @@ -22,6 +22,7 @@ use crate::admin::runtime_sources::{ current_object_store_handle_for_context, current_or_init_kms_runtime_service_manager, }; use crate::admin::storage_api::config::{read_admin_config, save_admin_config}; +use crate::admin::storage_api::error::StorageError; use crate::auth::{check_key_valid, get_session_token}; use crate::server::{ADMIN_PREFIX, RemoteAddr}; use hyper::{Method, StatusCode}; @@ -278,8 +279,11 @@ pub async fn load_kms_config() -> Option { } }, Err(e) => { - // Config not found is normal on first run - if e.to_string().contains("ConfigNotFound") || e.to_string().contains("not found") { + // Config not found is normal on first run: `read_config` maps a missing or + // empty config object to `ConfigNotFound`, so that variant is the only + // "absent" signal reaching here. Every other not-found variant (disk, + // volume, bucket) means degraded storage and must stay a warning. + if matches!(e, StorageError::ConfigNotFound) { info!( component = LOG_COMPONENT_ADMIN, subsystem = LOG_SUBSYSTEM_KMS, From 0f30a75fdb808baa6fe8fa5ac2a5f91bc42d0ab9 Mon Sep 17 00:00:00 2001 From: cxymds Date: Tue, 18 Aug 2026 14:49:19 +0800 Subject: [PATCH 06/24] fix(ecstore): resume remote shard reads once (#6091) * fix(ecstore): preserve CopyObject producer errors * fix(ecstore): resume remote shard reads once * fix(app): resume preserved relocation I/O errors * fix(ecstore): reserve remote read recovery budget --- .../cluster/rpc/internode_data_transport.rs | 25 + crates/ecstore/src/cluster/rpc/remote_disk.rs | 1187 ++++++++++++++++- crates/rio/src/http_reader.rs | 62 +- 3 files changed, 1258 insertions(+), 16 deletions(-) diff --git a/crates/ecstore/src/cluster/rpc/internode_data_transport.rs b/crates/ecstore/src/cluster/rpc/internode_data_transport.rs index 9c33d191a..6f46fead3 100644 --- a/crates/ecstore/src/cluster/rpc/internode_data_transport.rs +++ b/crates/ecstore/src/cluster/rpc/internode_data_transport.rs @@ -233,11 +233,17 @@ pub struct NsScannerCapabilityRequest { #[async_trait] pub trait InternodeDataTransport: Send + Sync + std::fmt::Debug { async fn open_read(&self, request: ReadStreamRequest) -> Result; + async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result { + self.open_read(request).await + } /// Opens an owned-chunk stream when this transport can retain receive-buffer /// ownership. `None` preserves the established `open_read` fallback. async fn open_read_chunks(&self, _request: ReadStreamRequest) -> Result> { Ok(None) } + async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result> { + self.open_read_chunks(request).await + } async fn open_write(&self, request: WriteStreamRequest) -> Result; async fn open_walk_dir(&self, request: WalkDirStreamRequest) -> Result; async fn open_ns_scanner(&self, _request: NsScannerStreamRequest) -> Result { @@ -269,6 +275,15 @@ impl InternodeDataTransport for TcpHttpInternodeDataTransport { )) } + async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result { + let url = build_read_file_stream_url(&request); + let mut headers = json_headers(); + build_auth_headers(&url, &Method::GET, &mut headers)?; + Ok(Box::new( + HttpReader::new_fresh_connection_with_stall_timeout(url, Method::GET, headers, None, request.stall_timeout).await?, + )) + } + async fn open_read_chunks(&self, request: ReadStreamRequest) -> Result> { let url = build_read_file_stream_url(&request); let mut headers = json_headers(); @@ -278,6 +293,16 @@ impl InternodeDataTransport for TcpHttpInternodeDataTransport { ))) } + async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result> { + let url = build_read_file_stream_url(&request); + let mut headers = json_headers(); + build_auth_headers(&url, &Method::GET, &mut headers)?; + Ok(Some(Box::new( + HttpChunkReader::new_fresh_connection_with_stall_timeout(url, Method::GET, headers, None, request.stall_timeout) + .await?, + ))) + } + async fn open_write(&self, request: WriteStreamRequest) -> Result { let server_epoch = self.put_file_auth_capability(&request.endpoint).await?; let nonce = server_epoch.map(|_| Uuid::new_v4()); diff --git a/crates/ecstore/src/cluster/rpc/remote_disk.rs b/crates/ecstore/src/cluster/rpc/remote_disk.rs index 13e45fbd3..3ae2c8b01 100644 --- a/crates/ecstore/src/cluster/rpc/remote_disk.rs +++ b/crates/ecstore/src/cluster/rpc/remote_disk.rs @@ -55,18 +55,22 @@ use rustfs_protos::proto_gen::node_service::{ }; use serde::{Serialize, de::DeserializeOwned}; use std::{ + future::Future, io::Cursor, path::PathBuf, + pin::Pin, sync::{ Arc, atomic::{AtomicU32, Ordering}, }, + task::{Context, Poll}, time::Duration, }; use tokio::time; use tokio::{ - io::{self, AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt}, + io::{self, AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt, ReadBuf}, net::TcpStream, + task::{JoinError, JoinHandle}, time::timeout, }; use tokio_util::sync::CancellationToken; @@ -84,6 +88,7 @@ const REMOTE_DISK_OPEN_WRITE_MAX_ATTEMPTS: usize = 2; const REMOTE_DISK_OPEN_WRITE_RETRY_BACKOFF: Duration = Duration::from_millis(20); const REMOTE_DISK_OPEN_READ_MAX_ATTEMPTS: usize = 2; const REMOTE_DISK_OPEN_READ_RETRY_BACKOFF: Duration = Duration::from_millis(20); +const REMOTE_READ_TIMEOUT_PARTS: u32 = 3; const NS_SCANNER_CAPABILITY_PROBE_TIMEOUT: Duration = Duration::from_secs(5); /// Base backoff for idempotent read-only RPC retries (grpc-optimization P3-3); doubles per attempt. const REMOTE_DISK_READ_RETRY_BASE_BACKOFF: Duration = Duration::from_millis(50); @@ -214,6 +219,415 @@ where } } +fn is_retryable_remote_body_error(error: &io::Error) -> bool { + if error + .get_ref() + .and_then(|source| source.downcast_ref::()) + .is_some() + { + return true; + } + + matches!( + error.kind(), + io::ErrorKind::ConnectionReset + | io::ErrorKind::BrokenPipe + | io::ErrorKind::ConnectionAborted + | io::ErrorKind::UnexpectedEof + ) +} + +fn resumed_read_request(request: &ReadStreamRequest, emitted: usize) -> io::Result { + let offset = request + .offset + .checked_add(emitted) + .ok_or_else(|| io::Error::other("remote read resume offset overflow"))?; + let length = if request.length == 0 { + 0 + } else { + request + .length + .checked_sub(emitted) + .ok_or_else(|| io::Error::other("remote read resume offset exceeds requested length"))? + }; + Ok(ReadStreamRequest { + offset, + length, + ..request.clone() + }) +} + +#[derive(Clone, Copy)] +struct RemoteReadTimeouts { + body_stall: Option, + initial_read: Option, + recovery: Option, +} + +fn remote_read_timeouts(read_timeout: Duration) -> RemoteReadTimeouts { + let Some(recovery) = read_timeout + .checked_div(REMOTE_READ_TIMEOUT_PARTS) + .filter(|timeout| !timeout.is_zero()) + else { + return RemoteReadTimeouts { + body_stall: None, + initial_read: None, + recovery: None, + }; + }; + RemoteReadTimeouts { + body_stall: Some(recovery), + initial_read: Some(read_timeout.saturating_sub(recovery)), + recovery: Some(recovery), + } +} + +async fn with_remote_read_recovery_timeout(recovery_timeout: Option, future: F) -> Result +where + F: Future>, +{ + match recovery_timeout { + Some(recovery_timeout) => match time::timeout(recovery_timeout, future).await { + Ok(result) => result, + Err(_) => Err(DiskError::Timeout), + }, + None => future.await, + } +} + +struct AbortOnDropTask(JoinHandle); + +impl AbortOnDropTask { + fn new(handle: JoinHandle) -> Self { + Self(handle) + } +} + +impl Future for AbortOnDropTask +where + T: Send + 'static, +{ + type Output = std::result::Result; + + fn poll(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll { + Pin::new(&mut self.get_mut().0).poll(cx) + } +} + +impl Drop for AbortOnDropTask { + fn drop(&mut self) { + self.0.abort(); + } +} + +fn retry_cutoff_elapsed( + initial_read_timeout: &mut Option, + cutoff: &mut Option>>, + cx: &mut Context<'_>, +) -> bool { + if cutoff.is_none() + && let Some(timeout) = initial_read_timeout.take() + { + *cutoff = Some(Box::pin(time::sleep(timeout))); + } + cutoff.as_mut().is_some_and(|cutoff| cutoff.as_mut().poll(cx).is_ready()) +} + +type ReadResumeFuture = AbortOnDropTask>; + +struct RetryingRemoteReader { + reader: Option, + transport: Arc, + request: ReadStreamRequest, + emitted: usize, + retried: bool, + initial_read_timeout: Option, + retry_cutoff: Option>>, + recovery_timeout: Option, + resume: Option, +} + +impl RetryingRemoteReader { + fn new_with_timeouts( + reader: FileReader, + transport: Arc, + request: ReadStreamRequest, + initial_read_timeout: Option, + recovery_timeout: Option, + ) -> Self { + Self { + reader: Some(reader), + transport, + request, + emitted: 0, + retried: false, + initial_read_timeout, + retry_cutoff: None, + recovery_timeout, + resume: None, + } + } + + fn start_resume(&mut self) -> io::Result<()> { + if self.request.length != 0 && self.emitted >= self.request.length { + self.reader = None; + return Ok(()); + } + let request = resumed_read_request(&self.request, self.emitted)?; + let recovery_timeout = self.recovery_timeout; + let transport = Arc::clone(&self.transport); + self.resume = Some(AbortOnDropTask::new(tokio::spawn(async move { + with_remote_read_recovery_timeout(recovery_timeout, transport.open_read_fresh(request)).await + }))); + Ok(()) + } + + fn retry_cutoff_elapsed(&mut self, cx: &mut Context<'_>) -> bool { + retry_cutoff_elapsed(&mut self.initial_read_timeout, &mut self.retry_cutoff, cx) + } +} + +impl AsyncRead for RetryingRemoteReader { + fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll> { + loop { + // After the absolute cutoff, let initial progress win over a stale fresh-open. + let resume_pending = if let Some(resume) = self.resume.as_mut() { + match Pin::new(resume).poll(cx) { + Poll::Pending => true, + Poll::Ready(Ok(Ok(reader))) => { + self.resume = None; + self.reader = Some(reader); + false + } + Poll::Ready(Ok(Err(error))) => { + self.resume = None; + if self.reader.is_none() { + return Poll::Ready(Err(io::Error::other(error))); + } + continue; + } + Poll::Ready(Err(error)) => { + self.resume = None; + if self.reader.is_none() { + return Poll::Ready(Err(io::Error::other(error))); + } + continue; + } + } + } else { + false + }; + + if !self.retried && self.retry_cutoff_elapsed(cx) { + self.retried = true; + if let Err(resume_error) = self.start_resume() { + return Poll::Ready(Err(resume_error)); + } + continue; + } + + let Some(reader) = self.reader.as_mut() else { + if resume_pending { + return Poll::Pending; + } + return Poll::Ready(Ok(())); + }; + let before = buf.filled().len(); + match Pin::new(reader).poll_read(cx, buf) { + Poll::Pending => return Poll::Pending, + Poll::Ready(Ok(())) => { + let produced = buf.filled().len() - before; + self.emitted = match self.emitted.checked_add(produced) { + Some(emitted) => emitted, + None => return Poll::Ready(Err(io::Error::other("remote read emitted byte count overflow"))), + }; + if resume_pending { + if produced == 0 && (self.request.length == 0 || self.emitted >= self.request.length) { + self.resume = None; + } else if produced == 0 { + self.reader = None; + continue; + } else { + self.resume = None; + } + } + return Poll::Ready(Ok(())); + } + Poll::Ready(Err(error)) if !self.retried && is_retryable_remote_body_error(&error) => { + self.retried = true; + self.reader = None; + if let Err(resume_error) = self.start_resume() { + return Poll::Ready(Err(resume_error)); + } + continue; + } + Poll::Ready(Err(error)) if resume_pending && is_retryable_remote_body_error(&error) => { + self.reader = None; + continue; + } + Poll::Ready(Err(error)) => return Poll::Ready(Err(error)), + } + } + } +} + +type ChunkResumeFuture = AbortOnDropTask>>; + +struct RetryingRemoteChunkReader { + reader: Option, + transport: Arc, + request: ReadStreamRequest, + emitted: usize, + retried: bool, + initial_read_timeout: Option, + retry_cutoff: Option>>, + recovery_timeout: Option, + resume: Option, +} + +impl RetryingRemoteChunkReader { + fn new_with_timeouts( + reader: rustfs_rio::ChunkReaderBox, + transport: Arc, + request: ReadStreamRequest, + initial_read_timeout: Option, + recovery_timeout: Option, + ) -> Self { + Self { + reader: Some(reader), + transport, + request, + emitted: 0, + retried: false, + initial_read_timeout, + retry_cutoff: None, + recovery_timeout, + resume: None, + } + } + + fn start_resume(&mut self) -> io::Result<()> { + if self.request.length != 0 && self.emitted >= self.request.length { + self.reader = None; + return Ok(()); + } + let request = resumed_read_request(&self.request, self.emitted)?; + let recovery_timeout = self.recovery_timeout; + let transport = Arc::clone(&self.transport); + self.resume = Some(AbortOnDropTask::new(tokio::spawn(async move { + with_remote_read_recovery_timeout(recovery_timeout, transport.open_read_chunks_fresh(request)).await + }))); + Ok(()) + } + + fn retry_cutoff_elapsed(&mut self, cx: &mut Context<'_>) -> bool { + retry_cutoff_elapsed(&mut self.initial_read_timeout, &mut self.retry_cutoff, cx) + } +} + +impl AsyncRead for RetryingRemoteChunkReader { + fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll> { + if buf.remaining() == 0 { + return Poll::Ready(Ok(())); + } + match rustfs_rio::ChunkReader::poll_read_chunk(self.as_mut(), cx, buf.remaining()) { + Poll::Ready(Ok(Some(chunk))) => { + buf.put_slice(&chunk); + Poll::Ready(Ok(())) + } + Poll::Ready(Ok(None)) => Poll::Ready(Ok(())), + Poll::Ready(Err(error)) => Poll::Ready(Err(error)), + Poll::Pending => Poll::Pending, + } + } +} + +impl rustfs_rio::ChunkReader for RetryingRemoteChunkReader { + fn poll_read_chunk(mut self: Pin<&mut Self>, cx: &mut Context<'_>, max: usize) -> Poll>> { + loop { + let resume_pending = if let Some(resume) = self.resume.as_mut() { + match Pin::new(resume).poll(cx) { + Poll::Pending => true, + Poll::Ready(Ok(Ok(Some(reader)))) => { + self.resume = None; + self.reader = Some(reader); + false + } + Poll::Ready(Ok(Ok(None))) => { + self.resume = None; + if self.reader.is_none() { + return Poll::Ready(Err(io::Error::other("remote resume transport did not provide a chunk reader"))); + } + continue; + } + Poll::Ready(Ok(Err(error))) => { + self.resume = None; + if self.reader.is_none() { + return Poll::Ready(Err(io::Error::other(error))); + } + continue; + } + Poll::Ready(Err(error)) => { + self.resume = None; + if self.reader.is_none() { + return Poll::Ready(Err(io::Error::other(error))); + } + continue; + } + } + } else { + false + }; + + if !self.retried && self.retry_cutoff_elapsed(cx) { + self.retried = true; + if let Err(resume_error) = self.start_resume() { + return Poll::Ready(Err(resume_error)); + } + continue; + } + + let Some(reader) = self.reader.as_mut() else { + if resume_pending { + return Poll::Pending; + } + return Poll::Ready(Ok(None)); + }; + match rustfs_rio::ChunkReader::poll_read_chunk(Pin::new(reader.as_mut()), cx, max) { + Poll::Pending => return Poll::Pending, + Poll::Ready(Ok(Some(chunk))) => { + self.emitted = match self.emitted.checked_add(chunk.len()) { + Some(emitted) => emitted, + None => return Poll::Ready(Err(io::Error::other("remote read emitted byte count overflow"))), + }; + if resume_pending { + self.resume = None; + } + return Poll::Ready(Ok(Some(chunk))); + } + Poll::Ready(Ok(None)) if resume_pending => { + self.reader = None; + continue; + } + Poll::Ready(Ok(None)) => return Poll::Ready(Ok(None)), + Poll::Ready(Err(error)) if !self.retried && is_retryable_remote_body_error(&error) => { + self.retried = true; + self.reader = None; + if let Err(resume_error) = self.start_resume() { + return Poll::Ready(Err(resume_error)); + } + continue; + } + Poll::Ready(Err(error)) if resume_pending && is_retryable_remote_body_error(&error) => { + self.reader = None; + continue; + } + Poll::Ready(Err(error)) => return Poll::Ready(Err(error)), + } + } + } +} + #[derive(Debug)] pub struct RemoteDisk { pub id: Mutex>, @@ -2483,17 +2897,24 @@ impl DiskAPI for RemoteDisk { return Err(DiskError::FaultyDisk); } let disk = self.disk_ref().await; - let stall_timeout = get_object_disk_read_timeout(); - self.open_read_with_retry(ReadStreamRequest { + let timeouts = remote_read_timeouts(get_object_disk_read_timeout()); + let request = ReadStreamRequest { endpoint: self.endpoint.grid_host(), disk, volume: volume.to_string(), path: path.to_string(), offset, length, - stall_timeout: (!stall_timeout.is_zero()).then_some(stall_timeout), - }) - .await + stall_timeout: timeouts.body_stall, + }; + let reader = self.open_read_with_retry(request.clone()).await?; + Ok(Box::new(RetryingRemoteReader::new_with_timeouts( + reader, + Arc::clone(&self.data_transport), + request, + timeouts.initial_read, + timeouts.recovery, + ))) } async fn read_file_stream_chunks( @@ -2507,17 +2928,26 @@ impl DiskAPI for RemoteDisk { return Err(DiskError::FaultyDisk); } let disk = self.disk_ref().await; - let stall_timeout = get_object_disk_read_timeout(); - self.open_read_chunks_with_retry(ReadStreamRequest { + let timeouts = remote_read_timeouts(get_object_disk_read_timeout()); + let request = ReadStreamRequest { endpoint: self.endpoint.grid_host(), disk, volume: volume.to_string(), path: path.to_string(), offset, length, - stall_timeout: (!stall_timeout.is_zero()).then_some(stall_timeout), - }) - .await + stall_timeout: timeouts.body_stall, + }; + let reader = self.open_read_chunks_with_retry(request.clone()).await?; + Ok(reader.map(|reader| { + Box::new(RetryingRemoteChunkReader::new_with_timeouts( + reader, + Arc::clone(&self.data_transport), + request, + timeouts.initial_read, + timeouts.recovery, + )) as rustfs_rio::ChunkReaderBox + })) } /// Buffered read for remote disks. @@ -3115,12 +3545,14 @@ impl DiskAPI for RemoteDisk { mod tests { use super::*; use crate::cluster::rpc::internode_data_transport::{InternodeDataTransportCapabilities, TcpHttpInternodeDataTransport}; + use crate::erasure::coding::{BitrotReader, Erasure, decode::ParallelReader}; + use crate::io_support::bitrot::ShardReader; use crate::runtime::sources as runtime_sources; use serde_json::Value; use serial_test::serial; use std::io::{self as std_io, Write}; use std::pin::Pin; - use std::sync::{Arc, Mutex, Mutex as StdMutex, Once}; + use std::sync::{Arc, Mutex, Mutex as StdMutex, Once, atomic::AtomicUsize}; use std::task::{Context, Poll}; use tokio::io::{ReadBuf, duplex}; use tokio::net::TcpListener; @@ -4138,6 +4570,735 @@ mod tests { } } + #[derive(Debug, Clone)] + enum ResumeReadStep { + PartialThenReset(Vec), + Data(Vec), + } + + #[derive(Debug, Default)] + struct ResumeTransport { + read_steps: Mutex>, + chunk_steps: Mutex>, + read_requests: Mutex>, + chunk_requests: Mutex>, + fresh_read_requests: Mutex>, + fresh_chunk_requests: Mutex>, + } + + impl ResumeTransport { + fn with_read_steps(read_steps: Vec) -> Self { + Self { + read_steps: Mutex::new(read_steps), + ..Self::default() + } + } + + fn with_chunk_steps(chunk_steps: Vec) -> Self { + Self { + chunk_steps: Mutex::new(chunk_steps), + ..Self::default() + } + } + } + + #[derive(Debug)] + struct ChunkPartialThenErrorReader { + data: Option, + error: Option, + } + + impl rustfs_rio::ChunkReader for ChunkPartialThenErrorReader { + fn poll_read_chunk(mut self: Pin<&mut Self>, _cx: &mut Context<'_>, max: usize) -> Poll>> { + if let Some(mut data) = self.data.take() { + let take = data.len().min(max); + let chunk = data.split_to(take); + if !data.is_empty() { + self.data = Some(data); + } + return Poll::Ready(Ok(Some(chunk))); + } + if let Some(error) = self.error.take() { + return Poll::Ready(Err(error)); + } + Poll::Ready(Ok(None)) + } + } + + impl AsyncRead for ChunkPartialThenErrorReader { + fn poll_read(self: Pin<&mut Self>, _cx: &mut Context<'_>, _buf: &mut ReadBuf<'_>) -> Poll> { + Poll::Ready(Err(io::Error::other("chunk reader must use chunk handoff"))) + } + } + + #[derive(Debug)] + struct BodyStallTestReader { + data: Option, + next_data: Option<(Bytes, Pin>)>, + initial_delay: Option>>, + timeout: Duration, + stall_timer: Option>>, + } + + impl BodyStallTestReader { + fn new(data: Bytes, initial_delay: Duration, timeout: Option) -> Self { + let timeout = timeout.expect("parallel resume test requires a body stall timeout"); + Self { + data: Some(data), + next_data: None, + initial_delay: Some(Box::pin(time::sleep(initial_delay))), + timeout, + stall_timer: None, + } + } + + fn with_next_data( + data: Bytes, + initial_delay: Duration, + next_data: Bytes, + next_delay: Duration, + timeout: Option, + ) -> Self { + let mut reader = Self::new(data, initial_delay, timeout); + reader.next_data = Some((next_data, Box::pin(time::sleep(next_delay)))); + reader + } + + fn poll_chunk(&mut self, cx: &mut Context<'_>, max: usize) -> Poll>> { + if let Some(delay) = self.initial_delay.as_mut() { + if delay.as_mut().poll(cx).is_pending() { + return Poll::Pending; + } + self.initial_delay = None; + } + if let Some(mut data) = self.data.take() { + let chunk = data.split_to(data.len().min(max)); + if !data.is_empty() { + self.data = Some(data); + } + return Poll::Ready(Ok(Some(chunk))); + } + if let Some((_, delay)) = self.next_data.as_mut() + && delay.as_mut().poll(cx).is_pending() + { + return Poll::Pending; + } + if let Some((mut data, _)) = self.next_data.take() { + let chunk = data.split_to(data.len().min(max)); + if !data.is_empty() { + self.next_data = Some((data, Box::pin(time::sleep(Duration::ZERO)))); + } + return Poll::Ready(Ok(Some(chunk))); + } + let timer = self.stall_timer.get_or_insert_with(|| Box::pin(time::sleep(self.timeout))); + match timer.as_mut().poll(cx) { + Poll::Pending => Poll::Pending, + Poll::Ready(()) => Poll::Ready(Err(io::Error::new( + std_io::ErrorKind::TimedOut, + rustfs_rio::BodyStalled { timeout: self.timeout }, + ))), + } + } + } + + impl AsyncRead for BodyStallTestReader { + fn poll_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll> { + match self.poll_chunk(cx, buf.remaining()) { + Poll::Ready(Ok(Some(chunk))) => { + buf.put_slice(&chunk); + Poll::Ready(Ok(())) + } + Poll::Ready(Ok(None)) => Poll::Ready(Ok(())), + Poll::Ready(Err(error)) => Poll::Ready(Err(error)), + Poll::Pending => Poll::Pending, + } + } + } + + impl rustfs_rio::ChunkReader for BodyStallTestReader { + fn poll_read_chunk(mut self: Pin<&mut Self>, cx: &mut Context<'_>, max: usize) -> Poll>> { + self.poll_chunk(cx, max) + } + } + + struct CountOnDrop(Arc); + + impl Drop for CountOnDrop { + fn drop(&mut self) { + self.0.fetch_add(1, Ordering::Relaxed); + } + } + + #[derive(Debug)] + struct ParallelResumeTransport { + initial_data: Bytes, + resumed_data: Bytes, + initial_delay: Duration, + fresh_delay: Duration, + fresh_read_requests: Mutex>, + fresh_chunk_requests: Mutex>, + initial_next_data: Option<(Bytes, Duration)>, + } + + impl ParallelResumeTransport { + fn new(initial_delay: Duration, fresh_delay: Duration) -> Self { + Self { + initial_data: Bytes::from_static(b"da"), + resumed_data: Bytes::from_static(b"ta"), + initial_delay, + fresh_delay, + fresh_read_requests: Mutex::new(Vec::new()), + fresh_chunk_requests: Mutex::new(Vec::new()), + initial_next_data: None, + } + } + + fn with_initial_next_data(initial_delay: Duration, next_delay: Duration, fresh_delay: Duration) -> Self { + let mut transport = Self::new(initial_delay, fresh_delay); + transport.initial_next_data = Some((Bytes::from_static(b"ta"), next_delay)); + transport + } + } + + #[async_trait::async_trait] + impl InternodeDataTransport for ParallelResumeTransport { + async fn open_read(&self, request: ReadStreamRequest) -> Result { + let reader = match self.initial_next_data.as_ref() { + Some((next_data, next_delay)) => BodyStallTestReader::with_next_data( + self.initial_data.clone(), + self.initial_delay, + next_data.clone(), + *next_delay, + request.stall_timeout, + ), + None => BodyStallTestReader::new(self.initial_data.clone(), self.initial_delay, request.stall_timeout), + }; + Ok(Box::new(reader)) + } + + async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result { + self.fresh_read_requests + .lock() + .expect("fresh read request lock should not be poisoned") + .push(request); + time::sleep(self.fresh_delay).await; + Ok(Box::new(Cursor::new(self.resumed_data.clone()))) + } + + async fn open_read_chunks(&self, request: ReadStreamRequest) -> Result> { + let reader = match self.initial_next_data.as_ref() { + Some((next_data, next_delay)) => BodyStallTestReader::with_next_data( + self.initial_data.clone(), + self.initial_delay, + next_data.clone(), + *next_delay, + request.stall_timeout, + ), + None => BodyStallTestReader::new(self.initial_data.clone(), self.initial_delay, request.stall_timeout), + }; + Ok(Some(Box::new(reader))) + } + + async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result> { + self.fresh_chunk_requests + .lock() + .expect("fresh chunk request lock should not be poisoned") + .push(request); + time::sleep(self.fresh_delay).await; + Ok(Some(Box::new(ChunkPartialThenErrorReader { + data: Some(self.resumed_data.clone()), + error: None, + }))) + } + + async fn open_write(&self, _request: WriteStreamRequest) -> Result { + panic!("open_write should not be used in parallel resume tests"); + } + + async fn open_walk_dir(&self, _request: WalkDirStreamRequest) -> Result { + panic!("open_walk_dir should not be used in parallel resume tests"); + } + + fn name(&self) -> &'static str { + "parallel-resume-test" + } + + fn capabilities(&self) -> InternodeDataTransportCapabilities { + InternodeDataTransportCapabilities::tcp_http() + } + } + + #[derive(Debug, Default)] + struct PendingFreshOpenTransport { + fresh_read_drops: Arc, + fresh_chunk_drops: Arc, + } + + #[async_trait::async_trait] + impl InternodeDataTransport for PendingFreshOpenTransport { + async fn open_read(&self, _request: ReadStreamRequest) -> Result { + Ok(Box::new(PartialThenErrorReader { + cursor: Cursor::new(Vec::new()), + error: Some(io::Error::new(std_io::ErrorKind::ConnectionReset, "stream reset")), + })) + } + + async fn open_read_fresh(&self, _request: ReadStreamRequest) -> Result { + let _drop = CountOnDrop(Arc::clone(&self.fresh_read_drops)); + std::future::pending().await + } + + async fn open_read_chunks(&self, _request: ReadStreamRequest) -> Result> { + Ok(Some(Box::new(ChunkPartialThenErrorReader { + data: None, + error: Some(io::Error::new(std_io::ErrorKind::ConnectionReset, "stream reset")), + }))) + } + + async fn open_read_chunks_fresh(&self, _request: ReadStreamRequest) -> Result> { + let _drop = CountOnDrop(Arc::clone(&self.fresh_chunk_drops)); + std::future::pending().await + } + + async fn open_write(&self, _request: WriteStreamRequest) -> Result { + panic!("open_write should not be used in fresh open cancellation tests"); + } + + async fn open_walk_dir(&self, _request: WalkDirStreamRequest) -> Result { + panic!("open_walk_dir should not be used in fresh open cancellation tests"); + } + + fn name(&self) -> &'static str { + "pending-fresh-open-test" + } + + fn capabilities(&self) -> InternodeDataTransportCapabilities { + InternodeDataTransportCapabilities::tcp_http() + } + } + + fn resume_step_reader(step: ResumeReadStep) -> FileReader { + match step { + ResumeReadStep::PartialThenReset(data) => Box::new(PartialThenErrorReader { + cursor: Cursor::new(data), + error: Some(io::Error::new(std_io::ErrorKind::ConnectionReset, "stream reset")), + }), + ResumeReadStep::Data(data) => Box::new(Cursor::new(data)), + } + } + + fn resume_step_chunk_reader(step: ResumeReadStep) -> rustfs_rio::ChunkReaderBox { + match step { + ResumeReadStep::PartialThenReset(data) => Box::new(ChunkPartialThenErrorReader { + data: Some(Bytes::from(data)), + error: Some(io::Error::new(std_io::ErrorKind::ConnectionReset, "stream reset")), + }), + ResumeReadStep::Data(data) => Box::new(ChunkPartialThenErrorReader { + data: Some(Bytes::from(data)), + error: None, + }), + } + } + + #[async_trait::async_trait] + impl InternodeDataTransport for ResumeTransport { + async fn open_read(&self, request: ReadStreamRequest) -> Result { + self.read_requests + .lock() + .expect("read request lock should not be poisoned") + .push(request); + let step = self + .read_steps + .lock() + .expect("read steps lock should not be poisoned") + .remove(0); + Ok(resume_step_reader(step)) + } + + async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result { + self.fresh_read_requests + .lock() + .expect("fresh read request lock should not be poisoned") + .push(request.clone()); + self.open_read(request).await + } + + async fn open_read_chunks(&self, request: ReadStreamRequest) -> Result> { + self.chunk_requests + .lock() + .expect("chunk request lock should not be poisoned") + .push(request); + let step = self + .chunk_steps + .lock() + .expect("chunk steps lock should not be poisoned") + .remove(0); + Ok(Some(resume_step_chunk_reader(step))) + } + + async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result> { + self.fresh_chunk_requests + .lock() + .expect("fresh chunk request lock should not be poisoned") + .push(request.clone()); + self.open_read_chunks(request).await + } + + async fn open_write(&self, _request: WriteStreamRequest) -> Result { + panic!("open_write should not be used in remote read resume tests"); + } + + async fn open_walk_dir(&self, _request: WalkDirStreamRequest) -> Result { + panic!("open_walk_dir should not be used in remote read resume tests"); + } + + fn name(&self) -> &'static str { + "resume-test" + } + + fn capabilities(&self) -> InternodeDataTransportCapabilities { + InternodeDataTransportCapabilities::tcp_http() + } + } + + fn resume_request(length: usize) -> ReadStreamRequest { + ReadStreamRequest { + endpoint: "http://remote".to_string(), + disk: "disk".to_string(), + volume: "volume".to_string(), + path: "path".to_string(), + offset: 7, + length, + stall_timeout: None, + } + } + + #[tokio::test] + async fn remote_reader_resumes_from_emitted_bytes_without_duplicates() { + let transport = Arc::new(ResumeTransport::with_read_steps(vec![ResumeReadStep::Data(b"456789".to_vec())])); + let request = resume_request(10); + let reader = resume_step_reader(ResumeReadStep::PartialThenReset(b"0123".to_vec())); + let mut reader = RetryingRemoteReader::new_with_timeouts(reader, transport.clone(), request, None, None); + let mut output = Vec::new(); + reader + .read_to_end(&mut output) + .await + .expect("one body reset should be resumed"); + + assert_eq!(output, b"0123456789"); + let requests = transport + .read_requests + .lock() + .expect("read request lock should not be poisoned"); + assert_eq!(requests.len(), 1); + assert_eq!(requests[0].offset, 11); + assert_eq!(requests[0].length, 6); + assert_eq!( + transport + .fresh_read_requests + .lock() + .expect("fresh read request lock should not be poisoned") + .len(), + 1 + ); + } + + #[tokio::test] + async fn remote_chunk_reader_resumes_from_emitted_bytes_without_duplicates() { + let transport = Arc::new(ResumeTransport::with_chunk_steps(vec![ResumeReadStep::Data(b"456789".to_vec())])); + let request = resume_request(10); + let reader = resume_step_chunk_reader(ResumeReadStep::PartialThenReset(b"0123".to_vec())); + let mut reader = RetryingRemoteChunkReader::new_with_timeouts(reader, transport.clone(), request, None, None); + let mut output = Vec::new(); + reader + .read_to_end(&mut output) + .await + .expect("chunk body reset should be resumed"); + + assert_eq!(output, b"0123456789"); + let requests = transport + .chunk_requests + .lock() + .expect("chunk request lock should not be poisoned"); + assert_eq!(requests.len(), 1); + assert_eq!(requests[0].offset, 11); + assert_eq!(requests[0].length, 6); + assert_eq!( + transport + .fresh_chunk_requests + .lock() + .expect("fresh chunk request lock should not be poisoned") + .len(), + 1 + ); + } + + #[derive(Clone, Copy)] + enum ParallelResumePath { + Regular, + Chunk, + } + + async fn assert_parallel_resume_case( + path: ParallelResumePath, + initial_delay: Duration, + initial_next_delay: Option, + fresh_delay: Duration, + expect_success: bool, + ) { + const DATA: &[u8] = b"data"; + let transport = Arc::new(match initial_next_delay { + Some(next_delay) => ParallelResumeTransport::with_initial_next_data(initial_delay, next_delay, fresh_delay), + None => ParallelResumeTransport::new(initial_delay, fresh_delay), + }); + let remote_disk = new_remote_disk_with_transport(transport.clone()).await; + let erasure = Erasure::new(1, 1, DATA.len()); + let (buffers, errors) = match path { + ParallelResumePath::Regular => { + let reader = remote_disk + .read_file_stream("bucket", "object/part.1", 0, DATA.len()) + .await + .expect("initial remote reader should open"); + let readers = vec![ + Some(BitrotReader::new(reader, DATA.len(), rustfs_utils::HashAlgorithm::None, false)), + None, + ]; + ParallelReader::new_with_metrics_path_and_reconstruction_verification(readers, erasure, 0, DATA.len(), None) + .read() + .await + } + ParallelResumePath::Chunk => { + let reader = remote_disk + .read_file_stream_chunks("bucket", "object/part.1", 0, DATA.len()) + .await + .expect("initial remote chunk reader should open") + .expect("chunk transport should return a reader"); + let readers = vec![ + Some(BitrotReader::new( + ShardReader::Chunked(reader), + DATA.len(), + rustfs_utils::HashAlgorithm::None, + false, + )), + None, + ]; + ParallelReader::new_with_metrics_path_and_reconstruction_verification(readers, erasure, 0, DATA.len(), None) + .read() + .await + } + }; + + if expect_success { + assert_eq!(buffers[0].as_deref(), Some(DATA)); + assert!(errors[0].is_none()); + } else { + assert!(buffers[0].is_none()); + assert!(matches!(errors[0], Some(DiskError::Timeout))); + } + let requests = match path { + ParallelResumePath::Regular => transport + .fresh_read_requests + .lock() + .expect("fresh read request lock should not be poisoned"), + ParallelResumePath::Chunk => transport + .fresh_chunk_requests + .lock() + .expect("fresh chunk request lock should not be poisoned"), + }; + assert_eq!(requests.len(), 1); + assert_eq!(requests[0].offset, 2); + assert_eq!(requests[0].length, 2); + } + + async fn assert_parallel_resume_path(path: ParallelResumePath) { + assert_parallel_resume_case(path, Duration::ZERO, None, Duration::from_millis(100), true).await; + assert_parallel_resume_case(path, Duration::from_millis(500), None, Duration::from_millis(100), true).await; + assert_parallel_resume_case(path, Duration::ZERO, None, Duration::from_millis(500), false).await; + assert_parallel_resume_case( + path, + Duration::from_millis(650), + Some(Duration::from_millis(700)), + Duration::from_millis(500), + true, + ) + .await; + } + + async fn assert_retry_drop_cancels_fresh_open(path: ParallelResumePath) { + let transport = Arc::new(PendingFreshOpenTransport::default()); + let remote_disk = new_remote_disk_with_transport(transport.clone()).await; + let mut output = [0_u8; 1]; + match path { + ParallelResumePath::Regular => { + let mut reader = remote_disk + .read_file_stream("bucket", "object/part.1", 0, 1) + .await + .expect("initial remote reader should open"); + assert!( + time::timeout(Duration::from_millis(20), reader.read(&mut output)) + .await + .is_err() + ); + drop(reader); + } + ParallelResumePath::Chunk => { + let mut reader = remote_disk + .read_file_stream_chunks("bucket", "object/part.1", 0, 1) + .await + .expect("initial remote chunk reader should open") + .expect("chunk transport should return a reader"); + assert!( + time::timeout(Duration::from_millis(20), reader.read(&mut output)) + .await + .is_err() + ); + drop(reader); + } + } + time::timeout(Duration::from_secs(1), async { + loop { + let drops = match path { + ParallelResumePath::Regular => transport.fresh_read_drops.load(Ordering::Relaxed), + ParallelResumePath::Chunk => transport.fresh_chunk_drops.load(Ordering::Relaxed), + }; + if drops == 1 { + break; + } + tokio::task::yield_now().await; + } + }) + .await + .expect("dropping the retrying reader should cancel the pending fresh open"); + } + + #[tokio::test(start_paused = true)] + #[serial] + async fn remote_reader_recovers_body_stall_through_parallel_reader() { + temp_env::async_with_vars([(rustfs_config::ENV_OBJECT_DISK_READ_TIMEOUT, Some("1"))], async { + assert_parallel_resume_path(ParallelResumePath::Regular).await; + }) + .await; + } + + #[tokio::test(start_paused = true)] + #[serial] + async fn remote_chunk_reader_recovers_body_stall_through_parallel_reader() { + temp_env::async_with_vars([(rustfs_config::ENV_OBJECT_DISK_READ_TIMEOUT, Some("1"))], async { + assert_parallel_resume_path(ParallelResumePath::Chunk).await; + }) + .await; + } + + #[tokio::test] + async fn remote_reader_drop_cancels_pending_fresh_open() { + assert_retry_drop_cancels_fresh_open(ParallelResumePath::Regular).await; + } + + #[tokio::test] + async fn remote_chunk_reader_drop_cancels_pending_fresh_open() { + assert_retry_drop_cancels_fresh_open(ParallelResumePath::Chunk).await; + } + + #[tokio::test] + async fn remote_reader_treats_error_after_requested_length_as_eof() { + let transport = Arc::new(ResumeTransport::default()); + let reader = resume_step_reader(ResumeReadStep::PartialThenReset(b"0123".to_vec())); + let mut reader = RetryingRemoteReader::new_with_timeouts(reader, transport.clone(), resume_request(4), None, None); + let mut output = Vec::new(); + + reader + .read_to_end(&mut output) + .await + .expect("error after the requested bytes should not trigger a redundant resume"); + assert_eq!(output, b"0123"); + assert!( + transport + .fresh_read_requests + .lock() + .expect("fresh read request lock should not be poisoned") + .is_empty() + ); + } + + #[tokio::test] + async fn remote_chunk_reader_treats_error_after_requested_length_as_eof() { + let transport = Arc::new(ResumeTransport::default()); + let reader = resume_step_chunk_reader(ResumeReadStep::PartialThenReset(b"0123".to_vec())); + let mut reader = RetryingRemoteChunkReader::new_with_timeouts(reader, transport.clone(), resume_request(4), None, None); + let mut output = Vec::new(); + + reader + .read_to_end(&mut output) + .await + .expect("error after the requested bytes should not trigger a redundant chunk resume"); + assert_eq!(output, b"0123"); + assert!( + transport + .fresh_chunk_requests + .lock() + .expect("fresh chunk request lock should not be poisoned") + .is_empty() + ); + } + + #[tokio::test] + async fn remote_reader_retries_at_most_once_and_preserves_non_retryable_errors() { + let transport = Arc::new(ResumeTransport::with_read_steps(vec![ResumeReadStep::PartialThenReset(b"456".to_vec())])); + let mut reader = RetryingRemoteReader::new_with_timeouts( + resume_step_reader(ResumeReadStep::PartialThenReset(b"0123".to_vec())), + transport.clone(), + resume_request(7), + None, + None, + ); + let error = reader + .read_to_end(&mut Vec::new()) + .await + .expect_err("second reset must not retry"); + assert_eq!(error.kind(), std_io::ErrorKind::ConnectionReset); + assert_eq!( + transport + .read_requests + .lock() + .expect("read request lock should not be poisoned") + .len(), + 1 + ); + + let transport = Arc::new(ResumeTransport::default()); + let reader = PartialThenErrorReader { + cursor: Cursor::new(b"data".to_vec()), + error: Some(io::Error::new(std_io::ErrorKind::PermissionDenied, "permission denied")), + }; + let mut reader = + RetryingRemoteReader::new_with_timeouts(Box::new(reader), transport.clone(), resume_request(4), None, None); + let error = reader + .read_to_end(&mut Vec::new()) + .await + .expect_err("non-retryable errors must not retry"); + assert_eq!(error.kind(), std_io::ErrorKind::PermissionDenied); + assert!( + transport + .read_requests + .lock() + .expect("read request lock should not be poisoned") + .is_empty() + ); + } + + #[test] + fn resumed_read_request_checks_large_offsets() { + let request = ReadStreamRequest { + offset: usize::MAX - 1, + length: 0, + ..resume_request(0) + }; + assert!(resumed_read_request(&request, 2).is_err()); + + let request = resume_request(4); + assert!(resumed_read_request(&request, 5).is_err()); + } + fn init_tracing(filter_level: Level) { INIT.call_once(|| { let _ = tracing_subscriber::fmt() @@ -4509,7 +5670,7 @@ mod tests { assert_eq!(request.path, "object/part.1"); assert_eq!(request.offset, 7); assert_eq!(request.length, 11); - assert_eq!(request.stall_timeout, Some(get_object_disk_read_timeout())); + assert_eq!(request.stall_timeout, remote_read_timeouts(get_object_disk_read_timeout()).body_stall); } other => panic!("expected read transport call, got {other:?}"), } diff --git a/crates/rio/src/http_reader.rs b/crates/rio/src/http_reader.rs index 96fd0971f..93f61a15c 100644 --- a/crates/rio/src/http_reader.rs +++ b/crates/rio/src/http_reader.rs @@ -715,6 +715,13 @@ async fn get_http_client(url: &str) -> io::Result { Ok(cached.client_for(disable_proxy)) } +async fn get_fresh_http_client(url: &str) -> io::Result { + let tuning = internode_http_client_tuning(); + let disable_proxy = should_disable_proxy_for_url(url, tuning); + let outbound_tls = crate::http_runtime_sources::outbound_tls_state().await; + build_http_client(disable_proxy, tuning, &outbound_tls).await +} + fn internode_request_context(method: &Method, url: &str, operation: Option<&'static str>) -> InternodeHttpRequestContext { let target = reqwest::Url::parse(url) .ok() @@ -962,6 +969,28 @@ impl HttpReader { Self::with_capacity_and_stall_timeout(url, method, headers, body, 0, stall_timeout).await } + pub async fn new_fresh_connection_with_stall_timeout( + url: String, + method: Method, + headers: HeaderMap, + body: Option>, + stall_timeout: Option, + ) -> io::Result { + let init = Self::open(&url, &method, &headers, body, stall_timeout, true).await?; + Ok(Self { + inner: StreamReader::new(init.stream), + url, + method, + headers, + track_internode_metrics: init.track_internode_metrics, + internode_operation: init.internode_operation, + stall_timer: None, + stall_timeout: init.stall_timeout, + request_started: init.request_started, + duration_recorded: false, + }) + } + /// Create a new HttpReader from a URL. The request is performed immediately. pub async fn with_capacity( url: String, @@ -981,7 +1010,7 @@ impl HttpReader { _read_buf_size: usize, stall_timeout: Option, ) -> io::Result { - let init = Self::open(&url, &method, &headers, body, stall_timeout).await?; + let init = Self::open(&url, &method, &headers, body, stall_timeout, false).await?; Ok(Self { inner: StreamReader::new(init.stream), url, @@ -1002,10 +1031,16 @@ impl HttpReader { headers: &HeaderMap, body: Option>, stall_timeout: Option, + force_fresh_connection: bool, ) -> io::Result { let track_internode_metrics = is_internode_rpc_url(url); let internode_operation = internode_rpc_operation(url); - let client = get_http_client(url).await.inspect_err(|_| { + let client = if force_fresh_connection { + get_fresh_http_client(url).await + } else { + get_http_client(url).await + } + .inspect_err(|_| { record_internode_error(track_internode_metrics, internode_operation); })?; let mut request: RequestBuilder = client.request(method.clone(), url).headers(headers.clone()); @@ -1121,7 +1156,28 @@ impl HttpChunkReader { body: Option>, stall_timeout: Option, ) -> io::Result { - let init = HttpReader::open(&url, &method, &headers, body, stall_timeout).await?; + let init = HttpReader::open(&url, &method, &headers, body, stall_timeout, false).await?; + Ok(Self { + inner: init.stream, + current: None, + track_internode_metrics: init.track_internode_metrics, + internode_operation: init.internode_operation, + stall_timer: None, + stall_timeout: init.stall_timeout, + request_started: init.request_started, + duration_recorded: false, + consecutive_empty_chunks: 0, + }) + } + + pub async fn new_fresh_connection_with_stall_timeout( + url: String, + method: Method, + headers: HeaderMap, + body: Option>, + stall_timeout: Option, + ) -> io::Result { + let init = HttpReader::open(&url, &method, &headers, body, stall_timeout, true).await?; Ok(Self { inner: init.stream, current: None, From 4f68f117ba485db1cff34fac79c3030f62723b69 Mon Sep 17 00:00:00 2001 From: cxymds Date: Tue, 18 Aug 2026 14:49:39 +0800 Subject: [PATCH 07/24] fix(lock): reap expired local lease guards (#6094) * fix(lock): reap expired local lease guards * fix(lock): reject refresh after guard expiry * test(lock): isolate expired refresh regression * style(lock): format expired refresh assertion --- crates/lock/src/client/local.rs | 351 +++++++++++++++++++++++------ crates/lock/src/namespace/tests.rs | 111 +++++++++ 2 files changed, 393 insertions(+), 69 deletions(-) diff --git a/crates/lock/src/client/local.rs b/crates/lock/src/client/local.rs index 2c65f2040..3b2a185aa 100644 --- a/crates/lock/src/client/local.rs +++ b/crates/lock/src/client/local.rs @@ -15,8 +15,10 @@ use std::collections::HashMap; use std::hash::{Hash, Hasher}; use std::sync::Arc; +use std::sync::atomic::{AtomicBool, Ordering}; use std::time::{Duration, SystemTime}; use tokio::sync::RwLock; +use tokio::time::Instant; use crate::{ FastLockGuard, GlobalLockManager, LockClient, LockId, LockInfo, LockManager, LockMetadata, LockPriority, LockRequest, @@ -26,43 +28,51 @@ use crate::{ /// Default shard count for guard storage (must be power of 2) const DEFAULT_GUARD_SHARD_COUNT: usize = 64; +type GuardShard = Arc>>; +type GuardStorage = Arc>; + /// Local lock client using FastLock with sharded guard storage for better concurrency #[derive(Debug)] pub struct LocalClient { /// Sharded guard storage to reduce lock contention - guard_storage: Vec>>>, + guard_storage: GuardStorage, /// Mask for fast shard index calculation (shard_count - 1) shard_mask: usize, /// Optional lock manager (if None, uses global singleton) manager: Option>, + reaper_started: AtomicBool, + reaper_interval: Duration, } #[derive(Debug)] struct LocalGuardEntry { guard: FastLockGuard, expires_at: SystemTime, + deadline: Instant, ttl: Duration, - /// Owner recorded at acquire time; used only for reclaim diagnostics (#899). - owner: String, } impl LocalGuardEntry { - fn new(guard: FastLockGuard, ttl: Duration, owner: String) -> Self { + fn new(guard: FastLockGuard, ttl: Duration) -> Self { let now = SystemTime::now(); + let monotonic_now = Instant::now(); Self { guard, - expires_at: now + ttl, + expires_at: now.checked_add(ttl).unwrap_or(now), + deadline: monotonic_now.checked_add(ttl).unwrap_or(monotonic_now), ttl, - owner, } } fn is_expired(&self) -> bool { - self.expires_at <= SystemTime::now() + self.deadline <= Instant::now() } fn refresh(&mut self) { - self.expires_at = SystemTime::now() + self.ttl; + let now = SystemTime::now(); + let monotonic_now = Instant::now(); + self.expires_at = now.checked_add(self.ttl).unwrap_or(now); + self.deadline = monotonic_now.checked_add(self.ttl).unwrap_or(monotonic_now); } } @@ -77,26 +87,38 @@ impl LocalClient { pub fn with_shard_count(shard_count: usize) -> Self { assert!(shard_count.is_power_of_two(), "Shard count must be power of 2"); - let guard_storage: Vec>>> = - (0..shard_count).map(|_| Arc::new(RwLock::new(HashMap::new()))).collect(); + let guard_storage: Vec = (0..shard_count).map(|_| Arc::new(RwLock::new(HashMap::new()))).collect(); + Self::with_storage(Arc::new(guard_storage), None, crate::fast_lock::CLEANUP_INTERVAL) + } + + fn with_storage(guard_storage: GuardStorage, manager: Option>, reaper_interval: Duration) -> Self { + let shard_count = guard_storage.len(); + debug_assert!(shard_count.is_power_of_two()); Self { guard_storage, shard_mask: shard_count - 1, - manager: None, + manager, + reaper_started: AtomicBool::new(false), + reaper_interval, } } /// Create new local client with a specific lock manager /// This allows simulating multi-node environments where each node has its own lock backend pub fn with_manager(manager: Arc) -> Self { - Self { - guard_storage: (0..DEFAULT_GUARD_SHARD_COUNT) - .map(|_| Arc::new(RwLock::new(HashMap::new()))) - .collect(), - shard_mask: DEFAULT_GUARD_SHARD_COUNT - 1, - manager: Some(manager), - } + let guard_storage = (0..DEFAULT_GUARD_SHARD_COUNT) + .map(|_| Arc::new(RwLock::new(HashMap::new()))) + .collect(); + Self::with_storage(Arc::new(guard_storage), Some(manager), crate::fast_lock::CLEANUP_INTERVAL) + } + + #[cfg(test)] + pub(crate) fn with_manager_and_reaper_interval(manager: Arc, reaper_interval: Duration) -> Self { + let guard_storage = (0..DEFAULT_GUARD_SHARD_COUNT) + .map(|_| Arc::new(RwLock::new(HashMap::new()))) + .collect(); + Self::with_storage(Arc::new(guard_storage), Some(manager), reaper_interval) } /// Get the lock manager (injected manager if available, otherwise global singleton) @@ -118,52 +140,63 @@ impl LocalClient { } async fn reclaim_expired_guards_for_resource(&self, resource: &crate::ObjectKey) -> usize { - let mut reclaimed = 0usize; + let expired_entries = Self::extract_expired_guards(&self.guard_storage, Some(resource)).await; + Self::release_reclaimed_guards(expired_entries, Some(resource)) + } - for shard in &self.guard_storage { - let expired_entries = { - let mut guards = shard.write().await; - let mut retained = HashMap::with_capacity(guards.len()); - let mut expired_entries = Vec::new(); + async fn extract_expired_guards(storage: &GuardStorage, resource: Option<&crate::ObjectKey>) -> Vec { + let mut expired_entries = Vec::new(); + for shard in storage.iter() { + let mut guards = shard.write().await; + expired_entries.extend( + guards + .extract_if(|lock_id, entry| { + resource.is_none_or(|resource| &lock_id.resource == resource) && entry.is_expired() + }) + .map(|(_, entry)| entry), + ); + } + expired_entries + } - for (lock_id, entry) in std::mem::take(&mut *guards) { - if &lock_id.resource == resource && entry.is_expired() { - expired_entries.push(entry); - } else { - retained.insert(lock_id, entry); - } - } - - *guards = retained; - expired_entries - }; - - for mut entry in expired_entries { - // An expired entry whose owner never refreshed it (a dead coordinator, #698) is - // reclaimed so a live contender can re-form quorum. With guard heartbeats in place - // (#899) a live owner keeps its entry from expiring, so reaching here means the - // lease genuinely lapsed. Surface it for observability; the reclaim decision itself - // is unchanged. - let since_last_refresh = entry - .expires_at - .checked_sub(entry.ttl) - .and_then(|last_refresh| SystemTime::now().duration_since(last_refresh).ok()) - .unwrap_or(entry.ttl); - tracing::warn!( - owner = %entry.owner, - resource = %resource, - ttl_ms = entry.ttl.as_millis() as u64, - since_last_refresh_ms = since_last_refresh.as_millis() as u64, - "reclaiming expired lock guard whose lease was not refreshed" - ); - rustfs_io_metrics::record_lock_reclaimed(); - let _ = entry.guard.release(); - reclaimed = reclaimed.saturating_add(1); + fn release_reclaimed_guards( + entries: impl IntoIterator, + resource: Option<&crate::ObjectKey>, + ) -> usize { + let mut reclaimed = 0; + for mut entry in entries { + let _ = entry.guard.release(); + rustfs_io_metrics::record_lock_reclaimed(); + reclaimed += 1; + } + if reclaimed > 0 { + if let Some(resource) = resource { + tracing::debug!(event = "lock_guard_reclaimed", resource = %resource, count = reclaimed, "expired lock guards reclaimed"); + } else { + tracing::debug!(event = "lock_guard_reaper_sweep", count = reclaimed, "expired lock guards reclaimed"); } } - reclaimed } + + fn ensure_reaper(&self) { + if self.reaper_started.swap(true, Ordering::AcqRel) { + return; + } + let storage = Arc::downgrade(&self.guard_storage); + let interval = self.reaper_interval; + tokio::spawn(async move { + let mut ticker = tokio::time::interval(interval); + loop { + ticker.tick().await; + let Some(storage) = storage.upgrade() else { + break; + }; + let expired_entries = Self::extract_expired_guards(&storage, None).await; + Self::release_reclaimed_guards(expired_entries, None); + } + }); + } } impl Default for LocalClient { @@ -175,28 +208,36 @@ impl Default for LocalClient { #[async_trait::async_trait] impl LockClient for LocalClient { async fn acquire_lock(&self, request: &LockRequest) -> Result { + self.ensure_reaper(); let lock_manager = self.get_lock_manager(); let reclaimed_before_acquire = self.reclaim_expired_guards_for_resource(&request.resource).await; + let acquire_deadline = Instant::now() + .checked_add(request.acquire_timeout) + .unwrap_or_else(Instant::now); - let build_lock_request = || match request.lock_type { + let build_lock_request = |acquire_timeout| match request.lock_type { LockType::Exclusive => crate::ObjectLockRequest::new_write(request.resource.clone(), request.owner.clone()) - .with_acquire_timeout(request.acquire_timeout), + .with_acquire_timeout(acquire_timeout), LockType::Shared => crate::ObjectLockRequest::new_read(request.resource.clone(), request.owner.clone()) - .with_acquire_timeout(request.acquire_timeout), + .with_acquire_timeout(acquire_timeout), }; let mut retried_after_reclaim = reclaimed_before_acquire > 0; loop { - match lock_manager.acquire_lock(build_lock_request()).await { + let remaining = acquire_deadline.saturating_duration_since(Instant::now()); + if remaining.is_zero() { + return Ok(LockResponse::failure("Lock acquisition timeout", request.acquire_timeout)); + } + match lock_manager.acquire_lock(build_lock_request(remaining)).await { Ok(guard) => { let lock_id = request.lock_id.clone(); let acquired_at = SystemTime::now(); - let expires_at = acquired_at + request.ttl; + let expires_at = acquired_at.checked_add(request.ttl).unwrap_or(acquired_at); { let shard = self.get_shard(&lock_id); let mut guards = shard.write().await; - guards.insert(lock_id.clone(), LocalGuardEntry::new(guard, request.ttl, request.owner.clone())); + guards.insert(lock_id.clone(), LocalGuardEntry::new(guard, request.ttl)); } let lock_info = LockInfo { @@ -256,12 +297,24 @@ impl LockClient for LocalClient { async fn refresh(&self, lock_id: &LockId) -> Result { let shard = self.get_shard(lock_id); - let mut guards = shard.write().await; - if let Some(entry) = guards.get_mut(lock_id) { - entry.refresh(); - Ok(true) - } else { + let expired_entry = { + let mut guards = shard.write().await; + let Some(entry) = guards.get_mut(lock_id) else { + return Ok(false); + }; + if entry.is_expired() { + guards.remove(lock_id) + } else { + entry.refresh(); + None + } + }; + + if let Some(entry) = expired_entry { + Self::release_reclaimed_guards([entry], Some(&lock_id.resource)); Ok(false) + } else { + Ok(true) } } @@ -317,3 +370,163 @@ impl LockClient for LocalClient { true } } + +#[cfg(test)] +mod tests { + use super::*; + use crate::{GlobalLockManager, LockClient, LockRequest, LockType}; + + fn request(resource: crate::ObjectKey, owner: &str, ttl: Duration) -> LockRequest { + LockRequest::new(resource, LockType::Exclusive, owner) + .with_ttl(ttl) + .with_acquire_timeout(Duration::from_millis(80)) + } + + async fn wait_until_reaped(client: &LocalClient, lock_id: &LockId) { + for _ in 0..80 { + if client.check_status(lock_id).await.unwrap().is_none() { + return; + } + tokio::time::sleep(Duration::from_millis(5)).await; + } + panic!("lock guard was not reaped before test deadline"); + } + + #[tokio::test(flavor = "current_thread")] + async fn expired_guard_is_reaped_without_resource_reacquire() { + let manager = Arc::new(GlobalLockManager::new()); + let client = LocalClient::with_manager_and_reaper_interval(manager.clone(), Duration::from_millis(5)); + let request = request(crate::ObjectKey::new("bucket", "unique-chunk"), "owner-a", Duration::from_millis(10)); + let lock_id = request.lock_id.clone(); + + assert!(client.acquire_lock(&request).await.unwrap().success); + assert!(client.check_status(&lock_id).await.unwrap().is_some()); + tokio::time::sleep(Duration::from_millis(15)).await; + wait_until_reaped(&client, &lock_id).await; + + let direct = manager + .acquire_lock(crate::ObjectLockRequest::new_write(request.resource.clone(), "owner-b")) + .await; + assert!(direct.is_ok()); + } + + #[tokio::test(flavor = "current_thread")] + async fn sibling_client_cannot_reclaim_but_owner_reaper_releases_shared_lock() { + let manager = Arc::new(GlobalLockManager::new()); + let owner = LocalClient::with_manager_and_reaper_interval(manager.clone(), Duration::from_millis(5)); + let contender = LocalClient::with_manager_and_reaper_interval(manager, Duration::from_millis(5)); + let request_a = request(crate::ObjectKey::new("bucket", "shared-resource"), "owner-a", Duration::from_millis(10)); + assert!(owner.acquire_lock(&request_a).await.unwrap().success); + + let request_b = request(request_a.resource.clone(), "owner-b", Duration::from_millis(20)) + .with_acquire_timeout(Duration::from_millis(5)); + assert!(!contender.acquire_lock(&request_b).await.unwrap().success); + + tokio::time::sleep(Duration::from_millis(25)).await; + assert!(owner.check_status(&request_a.lock_id).await.unwrap().is_none()); + assert!(contender.acquire_lock(&request_b).await.unwrap().success); + } + + #[tokio::test(flavor = "current_thread")] + async fn refresh_wins_before_deadline_and_reaper_wins_after_deadline() { + let manager = Arc::new(GlobalLockManager::new()); + let client = LocalClient::with_manager_and_reaper_interval(manager, Duration::from_millis(5)); + let request = request(crate::ObjectKey::new("bucket", "refresh-race"), "owner-a", Duration::from_millis(25)); + let lock_id = request.lock_id.clone(); + assert!(client.acquire_lock(&request).await.unwrap().success); + + tokio::time::sleep(Duration::from_millis(10)).await; + assert!(client.refresh(&lock_id).await.unwrap()); + tokio::time::sleep(Duration::from_millis(15)).await; + assert!(client.check_status(&lock_id).await.unwrap().is_some()); + wait_until_reaped(&client, &lock_id).await; + } + + #[tokio::test(start_paused = true)] + async fn refresh_after_expiry_releases_guard_without_reviving_it() { + let manager = Arc::new(GlobalLockManager::new()); + let client = LocalClient::with_manager_and_reaper_interval(manager, Duration::from_secs(60)); + client.reaper_started.store(true, Ordering::Release); + let lock_request = request( + crate::ObjectKey::new("bucket", "refresh-after-expiry"), + "owner-a", + Duration::from_secs(10), + ); + let lock_id = lock_request.lock_id.clone(); + + assert!( + client + .acquire_lock(&lock_request) + .await + .expect("initial owner should acquire the lock") + .success + ); + tokio::time::advance(Duration::from_secs(11)).await; + + assert!( + !client + .refresh(&lock_id) + .await + .expect("expired refresh should return a result"), + "an expired guard must not be refreshed" + ); + assert!( + client + .check_status(&lock_id) + .await + .expect("expired guard status should be readable") + .is_none(), + "expired guard should be removed after refresh" + ); + + let contender = request( + crate::ObjectKey::new("bucket", "refresh-after-expiry"), + "owner-b", + Duration::from_secs(10), + ); + assert!( + client + .acquire_lock(&contender) + .await + .expect("contender should receive an acquisition result") + .success, + "released guard must be acquirable by a new owner" + ); + } + + #[tokio::test(flavor = "current_thread")] + async fn zero_ttl_is_reaped_and_oversized_ttl_does_not_panic() { + let manager = Arc::new(GlobalLockManager::new()); + let client = LocalClient::with_manager_and_reaper_interval(manager, Duration::from_millis(5)); + + let zero = request(crate::ObjectKey::new("bucket", "zero-ttl"), "owner-zero", Duration::ZERO); + let zero_id = zero.lock_id.clone(); + assert!(client.acquire_lock(&zero).await.unwrap().success); + wait_until_reaped(&client, &zero_id).await; + + let huge = request(crate::ObjectKey::new("bucket", "huge-ttl"), "owner-huge", Duration::MAX); + let huge_id = huge.lock_id.clone(); + assert!(client.acquire_lock(&huge).await.unwrap().success); + wait_until_reaped(&client, &huge_id).await; + } + + #[tokio::test(flavor = "current_thread")] + async fn acquire_retry_preserves_total_deadline() { + let manager = Arc::new(GlobalLockManager::new()); + let client = LocalClient::with_manager_and_reaper_interval(manager, Duration::from_secs(60)); + let first = request(crate::ObjectKey::new("bucket", "deadline-budget"), "owner-a", Duration::from_millis(10)); + assert!(client.acquire_lock(&first).await.unwrap().success); + + let second = + request(first.resource.clone(), "owner-b", Duration::from_millis(30)).with_acquire_timeout(Duration::from_millis(60)); + let started = Instant::now(); + let response = client.acquire_lock(&second).await.unwrap(); + assert!(!response.success, "the first attempt consumed the caller's acquire budget"); + assert!( + started.elapsed() < Duration::from_millis(100), + "reclaim retry must not double the acquire budget" + ); + let recovered = client.acquire_lock(&second).await.unwrap(); + assert!(recovered.success, "the reclaimed guard must be available to the next request"); + } +} diff --git a/crates/lock/src/namespace/tests.rs b/crates/lock/src/namespace/tests.rs index 0ee5a81bb..3316828f7 100644 --- a/crates/lock/src/namespace/tests.rs +++ b/crates/lock/src/namespace/tests.rs @@ -840,6 +840,117 @@ async fn test_namespace_lock_distributed_reclaims_expired_same_resource_after_fa ); } +#[tokio::test] +async fn four_node_failed_release_converges_without_replica_repair() { + let managers = (0..4).map(|_| Arc::new(GlobalLockManager::new())).collect::>(); + let flaky_clients = managers + .iter() + .map(|manager| { + Arc::new(FlakyReleaseClient { + inner: LocalClient::with_manager_and_reaper_interval(manager.clone(), Duration::from_millis(5)), + failed_releases_remaining: AtomicUsize::new(usize::MAX), + release_attempts: AtomicUsize::new(0), + }) + }) + .collect::>(); + let clients = flaky_clients + .iter() + .map(|client| client.clone() as Arc) + .collect::>(); + let lock = NamespaceLock::Distributed(DistributedLock::new("four-node-expired-lease".to_string(), clients, 3)); + let resource = create_test_object_key("bucket", "object-four-node-expired"); + let request = LockRequest::new(resource.clone(), LockType::Exclusive, "owner-a") + .with_acquire_timeout(Duration::from_millis(300)) + .with_ttl(Duration::from_millis(40)); + + let mut guard = lock + .acquire_guard(&request) + .await + .expect("initial acquire should not error") + .expect("initial acquire should reach quorum"); + assert!(guard.release(), "release should be acknowledged while RPC cleanup is pending"); + + for _ in 0..40 { + if flaky_clients.iter().all(|client| client.release_attempts() >= 3) { + break; + } + tokio::time::sleep(Duration::from_millis(5)).await; + } + + let deadline = tokio::time::Instant::now() + Duration::from_secs(2); + loop { + let all_reaped = + futures::future::join_all(flaky_clients.iter().map(|client| client.inner.check_status(&request.lock_id))) + .await + .into_iter() + .all(|status| status.expect("status should not error").is_none()); + if all_reaped { + break; + } + assert!(tokio::time::Instant::now() < deadline, "all four local lease entries must converge"); + tokio::time::sleep(Duration::from_millis(10)).await; + } + + for suffix in ["chunk-0", "chunk-1", ".rustfs.sys/multipart/upload-0"] { + for client in &flaky_clients { + let orphan = LockRequest::new(create_test_object_key("bucket", suffix), LockType::Exclusive, "orphan") + .with_ttl(Duration::from_millis(25)); + assert!(client.inner.acquire_lock(&orphan).await.expect("orphan acquire").success); + } + } + tokio::time::sleep(Duration::from_millis(80)).await; + + let recovered = lock + .acquire_guard( + &LockRequest::new(resource, LockType::Exclusive, "owner-b") + .with_acquire_timeout(Duration::from_millis(300)) + .with_ttl(Duration::from_millis(40)), + ) + .await + .expect("recovery acquire should not error") + .expect("four-node quorum should recover after local reapers run"); + drop(recovered); +} + +#[tokio::test] +async fn four_node_stale_quorum_contention_respects_acquire_deadline() { + let managers = (0..4).map(|_| Arc::new(GlobalLockManager::new())).collect::>(); + let node_clients = managers + .iter() + .map(|manager| Arc::new(LocalClient::with_manager_and_reaper_interval(manager.clone(), Duration::from_millis(5)))) + .collect::>(); + let resource = create_test_object_key("bucket", "stale-quorum"); + let stale = LockRequest::new(resource.clone(), LockType::Exclusive, "stale-owner").with_ttl(Duration::from_millis(180)); + for client in &node_clients { + assert!(client.acquire_lock(&stale).await.expect("stale acquire").success); + } + + let clients = node_clients + .iter() + .map(|client| client.clone() as Arc) + .collect::>(); + let lock = NamespaceLock::Distributed(DistributedLock::new("stale-quorum-deadline".to_string(), clients, 3)); + let contender = LockRequest::new(resource.clone(), LockType::Exclusive, "new-owner") + .with_acquire_timeout(Duration::from_millis(150)) + .with_ttl(Duration::from_millis(100)); + let started = tokio::time::Instant::now(); + let response = lock.acquire_guard(&contender).await.expect("contention should not error"); + assert!(response.is_none(), "unexpired leases must not be force-reclaimed"); + assert!(started.elapsed() < Duration::from_millis(350), "acquire must respect its deadline"); + + tokio::time::sleep(Duration::from_millis(80)).await; + let recovered = lock + .acquire_guard( + &LockRequest::new(resource, LockType::Exclusive, "new-owner") + .with_acquire_timeout(Duration::from_millis(300)) + .with_ttl(Duration::from_millis(100)), + ) + .await + .expect("post-expiry acquire should not error") + .expect("quorum should recover after local reapers clear stale leases"); + drop(recovered); +} + #[tokio::test] async fn test_namespace_lock_distributed_retries_transient_acquire_timeout() { let managers = (0..3).map(|_| Arc::new(GlobalLockManager::new())).collect::>(); From a4ea36b298b5e1713c29178f03f8b6bdc796a761 Mon Sep 17 00:00:00 2001 From: cxymds Date: Tue, 18 Aug 2026 14:49:53 +0800 Subject: [PATCH 08/24] fix(ecstore): single-flight remote disk recovery (#6096) * fix(ecstore): single-flight remote disk recovery * test(ecstore): cover remote recovery review cases * test(ecstore): exercise recovery through disk slot * test(ecstore): match format reads exactly * test(ecstore): cover recovery teardown races * style(ecstore): format recovery race tests * fix(ecstore): remove unused health snapshot helper * fix(ecstore): group recovery monitor test state * fix(protos): preserve production source in compatibility checks --- Cargo.lock | 1 + crates/ecstore/Cargo.toml | 1 + crates/ecstore/src/cluster/rpc/remote_disk.rs | 661 +++++++++++++++++- crates/ecstore/src/disk/disk_store.rs | 130 +++- crates/protos/src/lib.rs | 2 +- 5 files changed, 746 insertions(+), 49 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index 82ffafc81..b297135af 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -9465,6 +9465,7 @@ dependencies = [ "tokio-stream", "tokio-util", "tonic", + "tonic-prost", "tower", "tracing", "tracing-core", diff --git a/crates/ecstore/Cargo.toml b/crates/ecstore/Cargo.toml index bdbb53e49..6525f0cfd 100644 --- a/crates/ecstore/Cargo.toml +++ b/crates/ecstore/Cargo.toml @@ -273,6 +273,7 @@ proptest = "1" rcgen.workspace = true insta = { workspace = true, features = ["yaml", "json"] } rustfs-crypto = { workspace = true } +tonic-prost = { workspace = true } [build-dependencies] shadow-rs = { workspace = true, default-features = false, features = ["build", "metadata"] } diff --git a/crates/ecstore/src/cluster/rpc/remote_disk.rs b/crates/ecstore/src/cluster/rpc/remote_disk.rs index 3ae2c8b01..26e9f899a 100644 --- a/crates/ecstore/src/cluster/rpc/remote_disk.rs +++ b/crates/ecstore/src/cluster/rpc/remote_disk.rs @@ -61,7 +61,7 @@ use std::{ pin::Pin, sync::{ Arc, - atomic::{AtomicU32, Ordering}, + atomic::{AtomicBool, AtomicU32, Ordering}, }, task::{Context, Poll}, time::Duration, @@ -630,6 +630,8 @@ impl rustfs_rio::ChunkReader for RetryingRemoteChunkReader { #[derive(Debug)] pub struct RemoteDisk { + /// Stable identity for this handle instance; replacement handles receive a new identity. + handle_id: Uuid, pub id: Mutex>, pub addr: String, endpoint: Endpoint, @@ -640,9 +642,38 @@ pub struct RemoteDisk { health: Arc, /// Cancellation token for monitoring tasks cancel_token: CancellationToken, + recovery_monitor_active: Arc, + #[cfg(test)] + recovery_monitor_start_count: Arc, + #[cfg(test)] + recovery_monitor_teardown_hook: Arc>>>, data_transport: Arc, } +struct RecoveryMonitorLease { + active: Arc, +} + +impl Drop for RecoveryMonitorLease { + fn drop(&mut self) { + self.active.store(false, Ordering::Release); + } +} + +#[cfg(test)] +#[derive(Debug, Default)] +struct RecoveryMonitorTeardownHook { + arrived: tokio::sync::Notify, + release: tokio::sync::Notify, +} + +#[cfg(test)] +#[derive(Clone)] +struct RecoveryMonitorTestState { + start_count: Arc, + teardown_hook: Arc>>>, +} + // ── Connection lifecycle (grpc-optimization P3) ── /// Whether to prewarm the internode control channel in the background at construction (default off). @@ -782,14 +813,15 @@ impl RemoteDisk { .await } - fn recovery_monitor_span(addr: &str, endpoint: &Endpoint) -> tracing::Span { + fn recovery_monitor_span(addr: &str, endpoint: &Endpoint, handle_id: Uuid) -> tracing::Span { tracing::info_span!( "recovery-monitor", component = LOG_COMPONENT_ECSTORE, subsystem = LOG_SUBSYSTEM_REMOTE_DISK, kind = "remote_disk", endpoint = %endpoint, - addr = %addr + addr = %addr, + handle_id = %handle_id ) } @@ -825,6 +857,7 @@ impl RemoteDisk { rustfs_utils::get_env_bool(ENV_RUSTFS_DRIVE_ACTIVE_MONITORING, DEFAULT_RUSTFS_DRIVE_ACTIVE_MONITORING); let disk = Self { + handle_id: Uuid::new_v4(), id: Mutex::new(None), addr, endpoint: ep.clone(), @@ -832,6 +865,11 @@ impl RemoteDisk { health_check: opt.health_check && env_health_check, health: Arc::new(DiskHealthTracker::new()), cancel_token: CancellationToken::new(), + recovery_monitor_active: Arc::new(AtomicBool::new(false)), + #[cfg(test)] + recovery_monitor_start_count: Arc::new(AtomicU32::new(0)), + #[cfg(test)] + recovery_monitor_teardown_hook: Arc::new(tokio::sync::Mutex::new(None)), data_transport, }; record_drive_runtime_state(ep, RuntimeDriveHealthState::Online); @@ -849,6 +887,16 @@ impl RemoteDisk { self.health.runtime_state() } + #[cfg(test)] + fn recovery_monitor_is_active(&self) -> bool { + self.recovery_monitor_active.load(Ordering::Acquire) + } + + #[cfg(test)] + fn recovery_monitor_start_count(&self) -> u32 { + self.recovery_monitor_start_count.load(Ordering::Acquire) + } + pub fn offline_duration_secs(&self) -> Option { self.health.offline_duration().map(|duration| duration.as_secs()) } @@ -987,13 +1035,62 @@ impl RemoteDisk { return; } - let addr = self.addr.clone(); - let endpoint = self.endpoint.clone(); - let health = Arc::clone(&self.health); - let cancel_token = self.cancel_token.clone(); - let span = Self::recovery_monitor_span(&addr, &endpoint); + Self::schedule_recovery_monitor( + self.addr.clone(), + self.endpoint.clone(), + self.handle_id, + Arc::clone(&self.health), + self.cancel_token.clone(), + Arc::clone(&self.recovery_monitor_active), + #[cfg(test)] + RecoveryMonitorTestState { + start_count: Arc::clone(&self.recovery_monitor_start_count), + teardown_hook: Arc::clone(&self.recovery_monitor_teardown_hook), + }, + ); + } + + fn schedule_recovery_monitor( + addr: String, + endpoint: Endpoint, + handle_id: Uuid, + health: Arc, + cancel_token: CancellationToken, + active: Arc, + #[cfg(test)] test_state: RecoveryMonitorTestState, + ) { + if active + .compare_exchange(false, true, Ordering::AcqRel, Ordering::Acquire) + .is_err() + { + return; + } + let span = Self::recovery_monitor_span(&addr, &endpoint, handle_id); super::spawn_background_monitor(span, async move { - Self::monitor_remote_disk_recovery(addr, endpoint, health, cancel_token).await; + #[cfg(test)] + test_state.start_count.fetch_add(1, Ordering::AcqRel); + let lease = RecoveryMonitorLease { + active: Arc::clone(&active), + }; + Self::monitor_remote_disk_recovery(addr.clone(), endpoint.clone(), Arc::clone(&health), cancel_token.clone()).await; + #[cfg(test)] + if let Some(hook) = test_state.teardown_hook.lock().await.take() { + hook.arrived.notify_one(); + hook.release.notified().await; + } + drop(lease); + if !cancel_token.is_cancelled() && health.runtime_state() != RuntimeDriveHealthState::Online { + Self::schedule_recovery_monitor( + addr, + endpoint, + handle_id, + health, + cancel_token, + active, + #[cfg(test)] + test_state, + ); + } }); } @@ -1002,7 +1099,7 @@ impl RemoteDisk { let (tx, rx) = tokio::sync::oneshot::channel(); let endpoint = self.endpoint.clone(); let addr = self.addr.clone(); - let span = Self::recovery_monitor_span(&addr, &endpoint); + let span = Self::recovery_monitor_span(&addr, &endpoint, self.handle_id); super::spawn_background_monitor(span, async move { warn!( event = EVENT_REMOTE_DISK_HEALTH, @@ -1033,9 +1130,23 @@ impl RemoteDisk { let cancel_token = self.cancel_token.clone(); let addr = self.addr.clone(); let endpoint = self.endpoint.clone(); + let handle_id = self.handle_id; + let recovery_monitor_active = Arc::clone(&self.recovery_monitor_active); + #[cfg(test)] + let recovery_monitor_teardown_hook = Arc::clone(&self.recovery_monitor_teardown_hook); tokio::spawn(async move { - Self::monitor_remote_disk_health(addr, endpoint, health, cancel_token).await; + Self::monitor_remote_disk_health( + addr, + endpoint, + handle_id, + health, + cancel_token, + recovery_monitor_active, + #[cfg(test)] + recovery_monitor_teardown_hook, + ) + .await; }); } @@ -1043,8 +1154,11 @@ impl RemoteDisk { async fn monitor_remote_disk_health( addr: String, endpoint: Endpoint, + handle_id: Uuid, health: Arc, cancel_token: CancellationToken, + recovery_monitor_active: Arc, + #[cfg(test)] recovery_monitor_teardown_hook: Arc>>>, ) { let mut interval = time::interval(get_drive_active_check_interval()); @@ -1069,11 +1183,19 @@ impl RemoteDisk { let addr_clone = addr.clone(); let endpoint_clone = endpoint.clone(); let cancel_clone = cancel_token.clone(); - let span = Self::recovery_monitor_span(&addr_clone, &endpoint_clone); - - super::spawn_background_monitor(span, async move { - Self::monitor_remote_disk_recovery(addr_clone, endpoint_clone, health_clone, cancel_clone).await; - }); + Self::schedule_recovery_monitor( + addr_clone, + endpoint_clone, + handle_id, + health_clone, + cancel_clone, + Arc::clone(&recovery_monitor_active), + #[cfg(test)] + RecoveryMonitorTestState { + start_count: Arc::new(AtomicU32::new(0)), + teardown_hook: Arc::clone(&recovery_monitor_teardown_hook), + }, + ); } loop { @@ -1132,11 +1254,19 @@ impl RemoteDisk { let addr_clone = addr.clone(); let endpoint_clone = endpoint.clone(); let cancel_clone = cancel_token.clone(); - let span = Self::recovery_monitor_span(&addr_clone, &endpoint_clone); - - super::spawn_background_monitor(span, async move { - Self::monitor_remote_disk_recovery(addr_clone, endpoint_clone, health_clone, cancel_clone).await; - }); + Self::schedule_recovery_monitor( + addr_clone, + endpoint_clone, + handle_id, + health_clone, + cancel_clone, + Arc::clone(&recovery_monitor_active), + #[cfg(test)] + RecoveryMonitorTestState { + start_count: Arc::new(AtomicU32::new(0)), + teardown_hook: Arc::clone(&recovery_monitor_teardown_hook), + }, + ); } } } @@ -1387,6 +1517,7 @@ impl RemoteDisk { subsystem = LOG_SUBSYSTEM_REMOTE_DISK, endpoint = %self.endpoint, addr = %self.addr, + handle_id = %self.handle_id, op, state = "faulty_short_circuit", "Remote disk operation short-circuited by faulty state" @@ -3548,15 +3679,23 @@ mod tests { use crate::erasure::coding::{BitrotReader, Erasure, decode::ParallelReader}; use crate::io_support::bitrot::ShardReader; use crate::runtime::sources as runtime_sources; + use rustfs_protos::proto_gen::node_service::{DiskInfoResponse, ReadAllResponse}; use serde_json::Value; use serial_test::serial; + use std::convert::Infallible; + use std::future::Future; use std::io::{self as std_io, Write}; use std::pin::Pin; use std::sync::{Arc, Mutex, Mutex as StdMutex, Once, atomic::AtomicUsize}; use std::task::{Context, Poll}; use tokio::io::{ReadBuf, duplex}; use tokio::net::TcpListener; - use tonic::transport::Endpoint as TonicEndpoint; + use tonic::transport::{Endpoint as TonicEndpoint, Server}; + use tonic::{Response, Status}; + use tonic::{ + codegen::{Body as HttpBody, BoxFuture, StdError, http}, + server::NamedService, + }; use tracing::Level; use tracing_subscriber::{Registry, fmt::MakeWriter, layer::SubscriberExt}; use uuid::Uuid; @@ -3716,6 +3855,218 @@ mod tests { ns_scanner_probe_status: Arc>>, } + #[derive(Clone, Debug)] + struct AuthenticatedReadPeer { + audience: String, + disk_info_calls: Arc, + read_all_calls: Arc, + object_read_all_disks: Arc>>, + format_data: Bytes, + read_all_data: Bytes, + } + + impl AuthenticatedReadPeer { + fn new(audience: String, format_data: Bytes, read_all_data: Bytes) -> Self { + Self { + audience, + disk_info_calls: Arc::new(AtomicU32::new(0)), + read_all_calls: Arc::new(AtomicU32::new(0)), + object_read_all_disks: Arc::default(), + format_data, + read_all_data, + } + } + + fn disk_info_calls(&self) -> u32 { + self.disk_info_calls.load(Ordering::Acquire) + } + + fn read_all_calls(&self) -> u32 { + self.read_all_calls.load(Ordering::Acquire) + } + + fn object_read_all_disks(&self) -> Vec { + self.object_read_all_disks + .lock() + .expect("object read_all disk list lock poisoned") + .clone() + } + + fn verify_auth(&self, request: &Request, path: &str) -> std::result::Result<(), Status> { + let headers = request.metadata().clone().into_headers(); + crate::cluster::rpc::verify_tonic_rpc_signature(&self.audience, path, &headers) + .map_err(|err| Status::unauthenticated(err.to_string())) + } + } + + #[derive(Clone, Debug)] + struct AuthenticatedReadPeerService { + peer: AuthenticatedReadPeer, + } + + impl NamedService for AuthenticatedReadPeerService { + const NAME: &'static str = "node_service.NodeService"; + } + + impl tower::Service> for AuthenticatedReadPeerService + where + B: HttpBody + Send + 'static, + B::Error: Into + Send + 'static, + { + type Response = http::Response; + type Error = Infallible; + type Future = BoxFuture; + + fn poll_ready(&mut self, _cx: &mut Context<'_>) -> Poll> { + Poll::Ready(Ok(())) + } + + fn call(&mut self, request: http::Request) -> Self::Future { + match request.uri().path() { + "/node_service.NodeService/DiskInfo" => { + #[derive(Clone)] + struct DiskInfoSvc(AuthenticatedReadPeer); + + impl tonic::server::UnaryService for DiskInfoSvc { + type Response = DiskInfoResponse; + type Future = Pin, Status>> + Send>>; + + fn call(&mut self, request: Request) -> Self::Future { + let peer = self.0.clone(); + Box::pin(async move { + peer.verify_auth(&request, "/node_service.NodeService/DiskInfo")?; + let request = request.into_inner(); + let opts = serde_json::from_str::(&request.opts) + .map_err(|err| Status::invalid_argument(err.to_string()))?; + if !opts.noop { + return Err(Status::invalid_argument("recovery probe must use noop disk_info")); + } + peer.disk_info_calls.fetch_add(1, Ordering::AcqRel); + let disk_info = serde_json::to_string(&DiskInfo { + total: 1, + free: 1, + endpoint: request.disk, + ..Default::default() + }) + .map_err(|err| Status::internal(err.to_string()))?; + Ok(Response::new(DiskInfoResponse { + success: true, + disk_info, + error: None, + })) + }) + } + } + + let peer = self.peer.clone(); + Box::pin(async move { + let method = DiskInfoSvc(peer); + let codec = tonic_prost::ProstCodec::default(); + let mut grpc = tonic::server::Grpc::new(codec); + Ok(grpc.unary(method, request).await) + }) + } + "/node_service.NodeService/ReadAll" => { + #[derive(Clone)] + struct ReadAllSvc(AuthenticatedReadPeer); + + impl tonic::server::UnaryService for ReadAllSvc { + type Response = ReadAllResponse; + type Future = Pin, Status>> + Send>>; + + fn call(&mut self, request: Request) -> Self::Future { + let peer = self.0.clone(); + Box::pin(async move { + peer.verify_auth(&request, "/node_service.NodeService/ReadAll")?; + let request = request.into_inner(); + let is_format_read = request.volume == crate::disk::RUSTFS_META_BUCKET + && request.path == crate::disk::FORMAT_CONFIG_FILE; + let disk = request.disk; + peer.read_all_calls.fetch_add(1, Ordering::AcqRel); + let data = if is_format_read { + peer.format_data.clone() + } else { + peer.object_read_all_disks + .lock() + .expect("object read_all disk list lock poisoned") + .push(disk); + peer.read_all_data.clone() + }; + Ok(Response::new(ReadAllResponse { + success: true, + data, + error: None, + })) + }) + } + } + + let peer = self.peer.clone(); + Box::pin(async move { + let method = ReadAllSvc(peer); + let codec = tonic_prost::ProstCodec::default(); + let mut grpc = tonic::server::Grpc::new(codec); + Ok(grpc.unary(method, request).await) + }) + } + _ => Box::pin(async move { + let mut response = http::Response::new(tonic::body::Body::default()); + let headers = response.headers_mut(); + headers.insert(tonic::Status::GRPC_STATUS, (tonic::Code::Unimplemented as i32).into()); + headers.insert(http::header::CONTENT_TYPE, tonic::metadata::GRPC_CONTENT_TYPE); + Ok(response) + }), + } + } + } + + struct TestGrpcPeer { + addr: String, + peer: AuthenticatedReadPeer, + shutdown: CancellationToken, + task: tokio::task::JoinHandle<()>, + } + + impl TestGrpcPeer { + async fn spawn(format_data: Bytes, read_all_data: Bytes) -> Option { + let listener = match TcpListener::bind("127.0.0.1:0").await { + Ok(listener) => listener, + Err(err) if err.kind() == std::io::ErrorKind::PermissionDenied => return None, + Err(err) => panic!("test gRPC listener should bind: {err}"), + }; + let socket_addr = listener.local_addr().expect("listener local address should be available"); + let addr = format!("http://{socket_addr}"); + let audience = crate::cluster::rpc::normalize_tonic_rpc_audience(&socket_addr.to_string()) + .expect("test audience should normalize"); + let peer = AuthenticatedReadPeer::new(audience, format_data, read_all_data); + let service = AuthenticatedReadPeerService { peer: peer.clone() }; + let shutdown = CancellationToken::new(); + let shutdown_for_task = shutdown.clone(); + let incoming = futures_util::stream::unfold(listener, |listener| async { + Some((listener.accept().await.map(|(stream, _)| stream), listener)) + }); + let task = tokio::spawn(async move { + Server::builder() + .add_service(service) + .serve_with_incoming_shutdown(incoming, shutdown_for_task.cancelled_owned()) + .await + .expect("test gRPC peer should serve"); + }); + + Some(Self { + addr, + peer, + shutdown, + task, + }) + } + + async fn stop(self) { + self.shutdown.cancel(); + let _ = self.task.await; + } + } + impl RecordingInternodeDataTransport { fn with_ns_scanner_probe_status(status: u16) -> Self { Self { @@ -5558,6 +5909,245 @@ mod tests { accept_task.abort(); } + #[tokio::test] + async fn faulty_handle_runs_only_one_recovery_monitor() { + let endpoint = Endpoint { + url: url::Url::parse("http://remote-node:9000/data/rustfs0").expect("endpoint should parse"), + is_local: false, + pool_idx: 0, + set_idx: 0, + disk_idx: 0, + }; + let disk = RemoteDisk::new( + &endpoint, + &DiskOption { + cleanup: false, + health_check: true, + }, + Arc::new(TcpHttpInternodeDataTransport), + ) + .await + .expect("remote disk should construct"); + if !disk.health_check { + return; + } + + disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline); + disk.spawn_recovery_monitor_if_needed(); + disk.spawn_recovery_monitor_if_needed(); + tokio::time::timeout(Duration::from_secs(1), async { + while disk.recovery_monitor_start_count() == 0 { + tokio::task::yield_now().await; + } + }) + .await + .expect("recovery monitor should start"); + + assert!(disk.recovery_monitor_is_active(), "only one recovery monitor should own the handle"); + assert_eq!( + disk.recovery_monitor_start_count(), + 1, + "the failed compare-exchange path must not start a second monitor" + ); + + disk.cancel_token.cancel(); + tokio::time::timeout(Duration::from_secs(1), async { + while disk.recovery_monitor_is_active() { + tokio::task::yield_now().await; + } + }) + .await + .expect("cancelled recovery monitor should release its single-flight state"); + assert!(!disk.recovery_monitor_is_active()); + } + + #[tokio::test] + #[serial(remote_disk_recovery_probe)] + async fn recovery_monitor_rearms_if_disk_fails_during_teardown() { + runtime_sources::ensure_test_rpc_secret(); + let Some(peer) = TestGrpcPeer::spawn(Bytes::new(), Bytes::new()).await else { + return; + }; + let endpoint = Endpoint { + url: url::Url::parse(&format!("{}/data/rustfs0", peer.addr)).expect("endpoint should parse"), + is_local: false, + pool_idx: 0, + set_idx: 0, + disk_idx: 0, + }; + let disk = RemoteDisk::new( + &endpoint, + &DiskOption { + cleanup: false, + health_check: true, + }, + Arc::new(TcpHttpInternodeDataTransport), + ) + .await + .expect("remote disk should construct"); + if !disk.health_check { + peer.stop().await; + return; + } + + disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline); + let hook = Arc::new(RecoveryMonitorTeardownHook::default()); + *disk.recovery_monitor_teardown_hook.lock().await = Some(Arc::clone(&hook)); + + temp_env::async_with_vars( + [ + (rustfs_config::ENV_DRIVE_RETURNING_PROBE_INTERVAL_SECS, Some("1")), + (rustfs_config::ENV_DRIVE_RETURNING_SUCCESS_THRESHOLD, Some("1")), + (rustfs_config::ENV_DRIVE_ACTIVE_CHECK_TIMEOUT_SECS, Some("1")), + ], + async { + disk.spawn_recovery_monitor_if_needed(); + tokio::time::timeout(Duration::from_secs(5), hook.arrived.notified()) + .await + .expect("first recovery monitor should reach teardown"); + assert_eq!(disk.runtime_state(), RuntimeDriveHealthState::Online); + + disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline); + hook.release.notify_one(); + tokio::time::timeout(Duration::from_secs(2), async { + while disk.recovery_monitor_start_count() < 2 { + tokio::task::yield_now().await; + } + }) + .await + .expect("teardown failure should re-arm recovery monitoring"); + assert!( + disk.recovery_monitor_is_active(), + "re-armed monitor should retain single-flight ownership" + ); + + disk.cancel_token.cancel(); + tokio::time::timeout(Duration::from_secs(2), async { + while disk.recovery_monitor_is_active() { + tokio::task::yield_now().await; + } + }) + .await + .expect("cancelled re-armed monitor should release single-flight state"); + }, + ) + .await; + + peer.stop().await; + } + + #[tokio::test] + #[serial(remote_disk_recovery_probe)] + async fn recovery_monitor_restores_online_then_real_reads_use_replacement_handle() { + runtime_sources::ensure_test_rpc_secret(); + let mut format = crate::layout::format::FormatV3::new(1, 1); + let disk_id = format.erasure.sets[0][0]; + format.erasure.this = disk_id; + let format_data = Bytes::from(format.to_json().expect("test format should serialize")); + let Some(peer) = TestGrpcPeer::spawn(format_data, Bytes::from_static(b"replacement-data")).await else { + return; + }; + let url = url::Url::parse(&format!("{}/data/rustfs0", peer.addr)).expect("endpoint should parse"); + let endpoint = Endpoint { + url, + is_local: false, + pool_idx: 0, + set_idx: 0, + disk_idx: 0, + }; + let disk = RemoteDisk::new( + &endpoint, + &DiskOption { + cleanup: false, + health_check: true, + }, + Arc::new(TcpHttpInternodeDataTransport), + ) + .await + .expect("remote disk should construct"); + disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline); + + temp_env::async_with_vars( + [ + (rustfs_config::ENV_DRIVE_RETURNING_PROBE_INTERVAL_SECS, Some("1")), + (rustfs_config::ENV_DRIVE_RETURNING_SUCCESS_THRESHOLD, Some("3")), + (rustfs_config::ENV_DRIVE_ACTIVE_CHECK_TIMEOUT_SECS, Some("1")), + ], + async { + let monitor = tokio::spawn(RemoteDisk::monitor_remote_disk_recovery( + disk.addr.clone(), + endpoint.clone(), + Arc::clone(&disk.health), + disk.cancel_token.clone(), + )); + + tokio::time::timeout(Duration::from_secs(5), async { + while disk.runtime_state() != RuntimeDriveHealthState::Online { + tokio::time::sleep(Duration::from_millis(20)).await; + } + }) + .await + .expect("three authenticated recovery probes should restore the disk online"); + monitor.await.expect("recovery monitor should exit after restoring Online"); + + assert_eq!( + peer.peer.disk_info_calls(), + 3, + "RemoteDisk recovery requires the configured three successful disk_info probes" + ); + let recovered_read = disk.read_all("bucket", "object").await.expect("recovered handle should read"); + assert_eq!(recovered_read, Bytes::from_static(b"replacement-data")); + + let old_disk = crate::disk::new_disk( + &endpoint, + &DiskOption { + cleanup: false, + health_check: false, + }, + ) + .await + .expect("old slot disk should construct"); + let set_disks = crate::set_disk::SetDisks::new( + "remote-recovery-test".to_string(), + Arc::new(tokio::sync::RwLock::new(vec![Some(old_disk.clone())])), + 1, + 0, + 0, + 0, + vec![endpoint.clone()], + format, + Vec::new(), + ) + .await; + set_disks.disks.write().await[0] = None; + set_disks.renew_disk(&endpoint).await; + + let slots = set_disks.disks.read().await; + let replacement = slots[0] + .as_ref() + .expect("renew_disk should publish the replacement slot") + .clone(); + drop(slots); + assert!(!Arc::ptr_eq(&replacement, &old_disk), "renew_disk must replace the stale slot handle"); + let replacement_read = replacement + .read_all("bucket", "object") + .await + .expect("production slot should route real reads through the replacement"); + assert_eq!(replacement_read, Bytes::from_static(b"replacement-data")); + let object_reads = peer.peer.object_read_all_disks(); + assert_eq!(object_reads.len(), 2, "standalone and production-slot reads should both reach the peer"); + assert_eq!(object_reads[1], disk_id.to_string(), "production slot must use the renewed disk identity"); + assert!(peer.peer.read_all_calls() >= 3, "renewal must read format metadata before the slot read"); + disk.cancel_token.cancel(); + old_disk.close().await.expect("old slot disk should close"); + replacement.close().await.expect("replacement slot disk should close"); + }, + ) + .await; + + peer.stop().await; + } + #[tokio::test] async fn test_copy_stream_with_buffer_copies_full_payload() { let payload = b"walk-dir-stream".repeat(1024); @@ -7124,6 +7714,20 @@ mod tests { ) .await .expect("remote disk should construct"); + let replacement = RemoteDisk::new( + &endpoint, + &DiskOption { + cleanup: false, + health_check: true, + }, + Arc::new(TcpHttpInternodeDataTransport), + ) + .await + .expect("replacement remote disk should construct"); + assert_ne!( + remote_disk.handle_id, replacement.handle_id, + "replacement handles need distinct log identities" + ); let span = tracing::info_span!("request-span", request_id = "req-remote-disk"); let _entered = span.enter(); @@ -7139,11 +7743,24 @@ mod tests { assert_eq!(log["span"]["name"], Value::String("recovery-monitor".to_string())); assert_eq!(log["span"]["kind"], Value::String("remote_disk".to_string())); + assert_eq!(log["span"]["handle_id"], Value::String(remote_disk.handle_id.to_string())); let spans = log["spans"].as_array().expect("spans should be present"); assert!(spans.iter().any(|span| { span.get("name").and_then(Value::as_str) == Some("request-span") && span.get("request_id").and_then(Value::as_str) == Some("req-remote-disk") })); + + remote_disk.force_runtime_state_for_test(RuntimeDriveHealthState::Offline); + remote_disk + .execute_with_timeout(|| async { Ok::<(), Error>(()) }, Duration::from_secs(1)) + .await + .expect_err("faulty handle should short-circuit"); + let faulty_log = logs + .lines() + .into_iter() + .find(|value| value.get("state").and_then(Value::as_str) == Some("faulty_short_circuit")) + .expect("expected faulty short-circuit log"); + assert_eq!(faulty_log["handle_id"], Value::String(remote_disk.handle_id.to_string())); } #[tokio::test(flavor = "current_thread")] diff --git a/crates/ecstore/src/disk/disk_store.rs b/crates/ecstore/src/disk/disk_store.rs index 29a8037dd..5c0f6454e 100644 --- a/crates/ecstore/src/disk/disk_store.rs +++ b/crates/ecstore/src/disk/disk_store.rs @@ -418,6 +418,17 @@ pub struct DiskHealthTracker { pub last_capacity_free: AtomicU64, /// Last successful capacity probe timestamp pub last_capacity_probe_unix_secs: AtomicI64, + /// Authoritative atomically published runtime/status pair. + state_snapshot: AtomicU64, + transition_lock: std::sync::Mutex<()>, +} + +fn pack_health_state(runtime_state: RuntimeDriveHealthState, status: u32) -> u64 { + (u64::from(runtime_state as u32) << 32) | u64::from(status) +} + +fn unpack_health_state(snapshot: u64) -> (RuntimeDriveHealthState, u32) { + (RuntimeDriveHealthState::from_u32((snapshot >> 32) as u32), snapshot as u32) } #[derive(Debug)] @@ -739,6 +750,8 @@ impl DiskHealthTracker { last_capacity_used: AtomicU64::new(0), last_capacity_free: AtomicU64::new(0), last_capacity_probe_unix_secs: AtomicI64::new(0), + state_snapshot: AtomicU64::new(pack_health_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK)), + transition_lock: std::sync::Mutex::new(()), } } @@ -775,39 +788,52 @@ impl DiskHealthTracker { /// Check if disk is faulty pub fn is_faulty(&self) -> bool { - self.status.load(Ordering::Acquire) == DISK_HEALTH_FAULTY + unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1 == DISK_HEALTH_FAULTY + } + + fn publish_state(&self, runtime_state: RuntimeDriveHealthState, status: u32) { + self.state_snapshot + .store(pack_health_state(runtime_state, status), Ordering::Release); + self.runtime_state.store(runtime_state as u32, Ordering::Release); + self.status.store(status, Ordering::Release); } /// Set disk as faulty #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")] pub fn set_faulty(&self) { - self.status.store(DISK_HEALTH_FAULTY, Ordering::Release); + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); + self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY); } /// Set disk as OK pub fn set_ok(&self) { - self.status.store(DISK_HEALTH_OK, Ordering::Release); + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); + self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK); } #[cfg(test)] pub fn force_runtime_state_for_test(&self, state: RuntimeDriveHealthState) { - self.runtime_state.store(state as u32, Ordering::Release); - match state { - RuntimeDriveHealthState::Offline => self.set_faulty(), - RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect | RuntimeDriveHealthState::Returning => { - self.set_ok(); - } - } + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); + let status = if state == RuntimeDriveHealthState::Offline { + DISK_HEALTH_FAULTY + } else { + DISK_HEALTH_OK + }; + self.publish_state(state, status); } pub fn swap_ok_to_faulty(&self) -> bool { - self.status - .compare_exchange(DISK_HEALTH_OK, DISK_HEALTH_FAULTY, Ordering::AcqRel, Ordering::Relaxed) - .is_ok() + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); + let (_, status) = unpack_health_state(self.state_snapshot.load(Ordering::Acquire)); + if status != DISK_HEALTH_OK { + return false; + } + self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY); + true } pub fn runtime_state(&self) -> RuntimeDriveHealthState { - RuntimeDriveHealthState::from_u32(self.runtime_state.load(Ordering::Acquire)) + unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).0 } pub fn offline_duration(&self) -> Option { @@ -823,6 +849,7 @@ impl DiskHealthTracker { } pub fn mark_failure(&self, endpoint: &Endpoint, reason: &'static str) -> bool { + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); let current = self.runtime_state(); let now = current_unix_secs(); let next = match current { @@ -851,24 +878,19 @@ impl DiskHealthTracker { }; let became_offline = next == RuntimeDriveHealthState::Offline && current != RuntimeDriveHealthState::Offline; - if next == RuntimeDriveHealthState::Offline { - self.status.store(DISK_HEALTH_FAULTY, Ordering::Release); - } else { - self.status.store(DISK_HEALTH_OK, Ordering::Release); - } self.transition_state(endpoint, current, next, reason); became_offline } #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")] pub fn mark_offline(&self, endpoint: &Endpoint, reason: &'static str) -> bool { + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); let current = self.runtime_state(); if current == RuntimeDriveHealthState::Offline { return false; } self.consecutive_successes.store(0, Ordering::Release); - self.status.store(DISK_HEALTH_FAULTY, Ordering::Release); self.transition_state(endpoint, current, RuntimeDriveHealthState::Offline, reason); true } @@ -882,11 +904,10 @@ impl DiskHealthTracker { } fn reset_for_store_init_retry_at(&self, endpoint: &Endpoint, now: Duration) { + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); let now_nanos = unix_nanos(now); let now_secs = unix_secs_i64(now); - self.status.store(DISK_HEALTH_OK, Ordering::Release); - self.runtime_state - .store(RuntimeDriveHealthState::Online as u32, Ordering::Release); + self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK); self.consecutive_failures.store(0, Ordering::Release); self.consecutive_successes.store(0, Ordering::Release); self.offline_since_unix_secs.store(0, Ordering::Release); @@ -898,6 +919,7 @@ impl DiskHealthTracker { } pub fn mark_recovery_success(&self, endpoint: &Endpoint, reason: &'static str) -> bool { + let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner()); let current = self.runtime_state(); let next = match current { RuntimeDriveHealthState::Online => RuntimeDriveHealthState::Online, @@ -918,7 +940,6 @@ impl DiskHealthTracker { let became_online = next == RuntimeDriveHealthState::Online; if became_online { - self.status.store(DISK_HEALTH_OK, Ordering::Release); self.consecutive_failures.store(0, Ordering::Release); self.consecutive_successes.store(0, Ordering::Release); } @@ -948,7 +969,13 @@ impl DiskHealthTracker { return; } - self.runtime_state.store(next as u32, Ordering::Release); + let current_status = unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1; + let status = match next { + RuntimeDriveHealthState::Offline => DISK_HEALTH_FAULTY, + RuntimeDriveHealthState::Returning => current_status, + RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect => DISK_HEALTH_OK, + }; + self.publish_state(next, status); self.last_transition_unix_secs .store(current_unix_secs() as i64, Ordering::Release); @@ -1217,7 +1244,7 @@ impl LocalDiskWrapper { return; } - if health.status.load(Ordering::Relaxed) != DISK_HEALTH_OK { + if health.is_faulty() { continue; } @@ -2909,6 +2936,57 @@ mod tests { }); } + #[test] + #[serial_test::serial] + fn concurrent_failure_and_recovery_publish_one_health_snapshot() { + temp_env::with_var(rustfs_config::ENV_DRIVE_SUSPECT_FAILURE_THRESHOLD, Some("2"), || { + let endpoint = Endpoint::try_from("/tmp/concurrent-health-snapshot").expect("endpoint should parse"); + let health = Arc::new(DiskHealthTracker::new()); + let transition_guard = health + .transition_lock + .lock() + .expect("health transition lock should not be poisoned"); + let start = Arc::new(std::sync::Barrier::new(3)); + let (completed_tx, completed_rx) = std::sync::mpsc::channel(); + let workers = (0..2) + .map(|_| { + let health = Arc::clone(&health); + let endpoint = endpoint.clone(); + let start = Arc::clone(&start); + let completed_tx = completed_tx.clone(); + std::thread::spawn(move || { + start.wait(); + health.mark_failure(&endpoint, "concurrent_test"); + completed_tx.send(()).expect("completion receiver should remain available"); + }) + }) + .collect::>(); + + start.wait(); + assert!( + matches!( + completed_rx.recv_timeout(Duration::from_millis(250)), + Err(std::sync::mpsc::RecvTimeoutError::Timeout) + ), + "concurrent transitions must wait for the serialization lock" + ); + drop(transition_guard); + completed_rx + .recv_timeout(Duration::from_secs(1)) + .expect("first failure transition should complete after lock release"); + completed_rx + .recv_timeout(Duration::from_secs(1)) + .expect("second failure transition should complete after lock release"); + for worker in workers { + worker.join().expect("health transition worker should not panic"); + } + + assert_eq!(health.runtime_state(), RuntimeDriveHealthState::Offline); + assert!(health.is_faulty()); + assert_eq!(health.consecutive_failures.load(Ordering::Acquire), 2); + }); + } + #[test] fn operation_success_recovers_suspect_drive_without_faulting() { let endpoint = Endpoint::try_from("/tmp/runtime-state-suspect-success").expect("endpoint should parse"); diff --git a/crates/protos/src/lib.rs b/crates/protos/src/lib.rs index a826d4a31..d23224b93 100644 --- a/crates/protos/src/lib.rs +++ b/crates/protos/src/lib.rs @@ -2557,7 +2557,7 @@ mod tests { fn production_source(source: &'static str, file_name: &str) -> &'static str { source - .split("\n#[cfg(test)]") + .split("\n#[cfg(test)]\nmod tests") .next() .unwrap_or_else(|| panic!("{file_name} should contain production source before tests")) } From a5800033bd323304d370bc81a31563b34222efdc Mon Sep 17 00:00:00 2001 From: houseme Date: Tue, 18 Aug 2026 16:09:30 +0800 Subject: [PATCH 09/24] feat(heal): incremental status cursors and typed overlap policy (HS-06) (#6206) * feat(heal): incremental heal status cursors and typed overlap policy (HS-06) Incremental results: every retained result item now carries a monotonic sequence number. The status query accepts a client cursor (sinceSeq on the admin wire, Option internally) and returns only newer items, plus nextSeq (the next cursor) and minSeq (the oldest retained sequence). A cursor that fell behind the 1024-item retention window is flagged through the existing truncated signal together with minSeq so the client can restart from it. Sequencing survives task completion: the completion archive stores the seq-stamped window. None keeps the exact legacy full-snapshot behavior, so existing clients see no change. Typed overlap handling for admin starts: RUSTFS_HEAL_OVERLAP_POLICY (merge default | minio_error). Under minio_error, an admin start whose path overlaps an active or queued task rejects with typed already-running / overlapping-paths admission reasons (surfaced through reason_label in the admin error body, sharing the existing OperationAborted site because the s3s footprint ratchet forbids new s3_error! sites); an exact duplicate start rejects with already-running instead of silently merging. Scanner/autoheal/ read-repair sources never take the rejection path. forceStart semantics now match MinIO for admin requests: an admin forceStart first cancels the overlapping active admin task, then admits the replacement. Wire: the heal-control Query command grows an optional sinceSeq (defaulted and skipped when absent, so older peers stay compatible); the admin handler accepts the sinceSeq query parameter; the local channel query gains the same cursor. Tests: seq monotonicity and incremental slicing, window slide moving minSeq with lagging-cursor flags, overlap matrix (same/containing/ contained/disjoint x policy x source), forceStart cancel-then-admit, and the completion-archive window handoff. Co-Authored-By: heihutu * style: fmt after main merge --------- Co-authored-by: heihutu Co-authored-by: zhi22915 --- Cargo.lock | 2 + crates/common/src/heal_channel.rs | 23 ++ crates/config/src/constants/heal.rs | 9 + crates/heal/src/heal/channel.rs | 113 +++++- crates/heal/src/heal/manager.rs | 497 ++++++++++++++++++++++--- crates/heal/src/heal/task.rs | 140 ++++++- crates/protos/src/heal_control.rs | 76 +++- crates/scanner/src/scanner_folder.rs | 7 + rustfs/src/admin/handlers/heal.rs | 24 +- rustfs/src/storage/rpc/node_service.rs | 10 +- scripts/check_s3s_footprint.sh | 2 +- 11 files changed, 807 insertions(+), 96 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index b297135af..83749d574 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -9539,6 +9539,8 @@ version = "1.0.0-rc.2" dependencies = [ "async-trait", "base64 0.23.1", + "bytes", + "crc-fast", "futures", "hotpath", "http 1.5.0", diff --git a/crates/common/src/heal_channel.rs b/crates/common/src/heal_channel.rs index b5a907e4e..f8162dee6 100644 --- a/crates/common/src/heal_channel.rs +++ b/crates/common/src/heal_channel.rs @@ -224,6 +224,13 @@ pub struct HealOpts { pub enum HealAdmissionDropReason { QueueFull, PolicyDropped, + /// HS-06: an admin heal start overlaps (same bucket with mutually + /// containing prefixes, or the same erasure set) an already running or + /// queued task. Only produced when RUSTFS_HEAL_OVERLAP_POLICY=minio_error. + AlreadyRunning, + /// HS-06: same as [`Self::AlreadyRunning`] but for paths that merely + /// contain (or are contained by) the active task's path. + OverlappingPaths, } impl HealAdmissionDropReason { @@ -231,6 +238,8 @@ impl HealAdmissionDropReason { match self { Self::QueueFull => "queue_full", Self::PolicyDropped => "policy_dropped", + Self::AlreadyRunning => "already_running", + Self::OverlappingPaths => "overlapping_paths", } } } @@ -317,6 +326,9 @@ pub enum HealChannelCommand { Query { heal_path: String, client_token: String, + /// Incremental result cursor (HS-06): only items with a sequence + /// greater than this are returned; `None` keeps the full snapshot. + since_seq: Option, response_tx: oneshot::Sender>, }, /// Cancel heal task @@ -522,10 +534,21 @@ async fn receive_heal_channel_response( /// Send heal query request pub async fn query_heal_status(heal_path: String, client_token: String) -> Result { + query_heal_status_since(heal_path, client_token, None).await +} + +/// Incremental heal query (HS-06): pass the client's last seen sequence +/// number to receive only newer result items. +pub async fn query_heal_status_since( + heal_path: String, + client_token: String, + since_seq: Option, +) -> Result { let (response_tx, response_rx) = oneshot::channel(); send_heal_command(HealChannelCommand::Query { heal_path, client_token, + since_seq, response_tx, }) .await?; diff --git a/crates/config/src/constants/heal.rs b/crates/config/src/constants/heal.rs index b8cf3630a..9c27a663f 100644 --- a/crates/config/src/constants/heal.rs +++ b/crates/config/src/constants/heal.rs @@ -205,3 +205,12 @@ pub const DEFAULT_HEAL_MRF_JOURNAL_MAX_BYTES: usize = 8 * 1024 * 1024; /// Default MRF replay batch size. pub const DEFAULT_HEAL_MRF_REPLAY_BATCH: usize = 256; + +/// Environment variable selecting how admin heal starts behave when the +/// requested path overlaps an already running or queued heal: `merge` +/// (default, keep today's dedup/merge semantics) or `minio_error` (return a +/// typed already-running / overlapping-paths rejection like madmin). +pub const ENV_HEAL_OVERLAP_POLICY: &str = "RUSTFS_HEAL_OVERLAP_POLICY"; + +/// Default overlap policy: merge duplicate/overlapping requests. +pub const DEFAULT_HEAL_OVERLAP_POLICY: &str = "merge"; diff --git a/crates/heal/src/heal/channel.rs b/crates/heal/src/heal/channel.rs index 1e31056b2..8f4ab9e8a 100644 --- a/crates/heal/src/heal/channel.rs +++ b/crates/heal/src/heal/channel.rs @@ -66,21 +66,37 @@ struct HealTaskStatusPayload<'a> { summary: &'a str, items: &'a [HealResultItem], truncated: bool, + /// Cursor for incremental consumption (HS-06): sequence of the next item + /// to be produced. Absent on responses without sequencing (0). + #[serde(skip_serializing_if = "u64_is_zero")] + next_seq: u64, + /// Oldest sequence still retained; with `truncated`, tells a lagging + /// client where to restart its cursor. + #[serde(skip_serializing_if = "u64_is_zero")] + min_seq: u64, #[serde(skip_serializing_if = "Option::is_none")] progress: Option<&'a HealProgress>, } +fn u64_is_zero(value: &u64) -> bool { + *value == 0 +} + fn encode_heal_task_status_payload( summary: &str, mut items: Vec, progress: Option<&HealProgress>, mut truncated: bool, + next_seq: u64, + min_seq: u64, ) -> Result<(Vec, bool)> { loop { let data = serde_json::to_vec(&HealTaskStatusPayload { summary, items: &items, truncated, + next_seq, + min_seq, progress, }) .map_err(|e| Error::Serialization(format!("failed to serialize heal task status: {e}")))?; @@ -109,8 +125,10 @@ fn encode_heal_status_response( progress: Option<&HealProgress>, detail: Option, truncated: bool, + next_seq: u64, + min_seq: u64, ) -> Result<(Vec, Option)> { - let (data, truncated) = encode_heal_task_status_payload(summary, items, progress, truncated)?; + let (data, truncated) = encode_heal_task_status_payload(summary, items, progress, truncated, next_seq, min_seq)?; Ok((data, heal_status_detail(detail, truncated))) } @@ -138,8 +156,19 @@ impl HealChannelProcessor { /// Execute a token query directly against the manager. pub async fn execute_query_request(&self, heal_path: String, client_token: String) -> Result { + self.execute_query_request_since(heal_path, client_token, None).await + } + + /// Incremental variant of [`Self::execute_query_request`] (HS-06). + pub async fn execute_query_request_since( + &self, + heal_path: String, + client_token: String, + since_seq: Option, + ) -> Result { let (response_tx, response_rx) = oneshot::channel(); - self.process_query_request(heal_path, client_token, response_tx).await?; + self.process_query_request(heal_path, client_token, since_seq, response_tx) + .await?; response_rx .await .map_err(|err| Error::other(format!("heal query channel closed: {err}")))? @@ -262,8 +291,12 @@ impl HealChannelProcessor { HealChannelCommand::Query { heal_path, client_token, + since_seq, response_tx, - } => self.process_query_request(heal_path, client_token, response_tx).await, + } => { + self.process_query_request(heal_path, client_token, since_seq, response_tx) + .await + } HealChannelCommand::Cancel { heal_path, client_token, @@ -384,6 +417,7 @@ impl HealChannelProcessor { &self, heal_path: String, client_token: String, + since_seq: Option, response_tx: oneshot::Sender>, ) -> Result<()> { debug!( @@ -398,72 +432,118 @@ impl HealChannelProcessor { ); let report = if heal_path.trim_matches('/').is_empty() { - self.heal_manager.get_task_report(&client_token).await + self.heal_manager.get_task_report_since(&client_token, since_seq).await } else { - self.heal_manager.get_task_report_for_path(&heal_path, &client_token).await + self.heal_manager + .get_task_report_for_path_since(&heal_path, &client_token, since_seq) + .await }; - let (summary, detail, items, truncated, progress) = match report { + let (summary, detail, items, truncated, progress, next_seq, min_seq) = match report { Ok(HealTaskReport { status: HealTaskStatus::Pending | HealTaskStatus::Running, result_items, result_items_truncated, progress, - }) => ("running".to_string(), None, result_items, result_items_truncated, progress), + next_seq, + min_seq, + }) => ( + "running".to_string(), + None, + result_items, + result_items_truncated, + progress, + next_seq, + min_seq, + ), Ok(HealTaskReport { status: HealTaskStatus::Retrying { error, retry_attempt }, result_items, result_items_truncated, progress, + next_seq, + min_seq, }) => ( "running".to_string(), Some(format!("heal task retrying after recoverable failure, attempt {retry_attempt}: {error}")), result_items, result_items_truncated, progress, + next_seq, + min_seq, ), Ok(HealTaskReport { status: HealTaskStatus::Completed, result_items, result_items_truncated, progress, - }) => ("finished".to_string(), None, result_items, result_items_truncated, progress), + next_seq, + min_seq, + }) => ( + "finished".to_string(), + None, + result_items, + result_items_truncated, + progress, + next_seq, + min_seq, + ), Ok(HealTaskReport { status: HealTaskStatus::Cancelled, result_items, result_items_truncated, progress, + next_seq, + min_seq, }) => ( "stopped".to_string(), Some("heal task cancelled".to_string()), result_items, result_items_truncated, progress, + next_seq, + min_seq, ), Ok(HealTaskReport { status: HealTaskStatus::Timeout, result_items, result_items_truncated, progress, + next_seq, + min_seq, }) => ( "stopped".to_string(), Some("heal task timed out".to_string()), result_items, result_items_truncated, progress, + next_seq, + min_seq, ), Ok(HealTaskReport { status: HealTaskStatus::Failed { error }, result_items, result_items_truncated, progress, - }) => ("stopped".to_string(), Some(error), result_items, result_items_truncated, progress), + next_seq, + min_seq, + }) => ( + "stopped".to_string(), + Some(error), + result_items, + result_items_truncated, + progress, + next_seq, + min_seq, + ), Err(crate::Error::TaskNotFound { .. }) => ( "notFound".to_string(), Some("heal task not found or expired".to_string()), Vec::new(), false, None, + 0, + 0, ), Err(crate::Error::InvalidClientToken) => { let response = HealChannelResponse { @@ -490,7 +570,8 @@ impl HealChannelProcessor { } }; - let (data, detail) = encode_heal_status_response(&summary, items, progress.as_ref(), detail, truncated)?; + let (data, detail) = + encode_heal_status_response(&summary, items, progress.as_ref(), detail, truncated, next_seq, min_seq)?; let response = HealChannelResponse { request_id: client_token, @@ -805,7 +886,7 @@ mod tests { ..Default::default() }]; - let (data, detail) = encode_heal_status_response("running", items, None, None, false).unwrap(); + let (data, detail) = encode_heal_status_response("running", items, None, None, false, 0, 0).unwrap(); assert!(data.len() <= MAX_HEAL_STATUS_PAYLOAD_SIZE); let payload: serde_json::Value = serde_json::from_slice(&data).unwrap(); @@ -1575,7 +1656,7 @@ mod tests { let (tx, rx) = oneshot::channel(); processor - .process_query_request("bucket".to_string(), "completed-token".to_string(), tx) + .process_query_request("bucket".to_string(), "completed-token".to_string(), None, tx) .await .expect("query should process"); @@ -1610,7 +1691,7 @@ mod tests { let (tx, rx) = oneshot::channel(); processor - .process_query_request("bucket".to_string(), task_id.clone(), tx) + .process_query_request("bucket".to_string(), task_id.clone(), None, tx) .await .expect("query should process"); @@ -1643,7 +1724,7 @@ mod tests { let (tx, rx) = oneshot::channel(); processor - .process_query_request("bucket".to_string(), "wrong-token".to_string(), tx) + .process_query_request("bucket".to_string(), "wrong-token".to_string(), None, tx) .await .expect("query should process"); @@ -1668,7 +1749,7 @@ mod tests { let (tx, rx) = oneshot::channel(); processor - .process_query_request(String::new(), "wrong-token".to_string(), tx) + .process_query_request(String::new(), "wrong-token".to_string(), None, tx) .await .expect("query should process"); @@ -1705,7 +1786,7 @@ mod tests { let (tx, rx) = oneshot::channel(); processor - .process_query_request(String::new(), task_id.clone(), tx) + .process_query_request(String::new(), task_id.clone(), None, tx) .await .expect("query should process"); diff --git a/crates/heal/src/heal/manager.rs b/crates/heal/src/heal/manager.rs index 216e17068..3a243c88e 100644 --- a/crates/heal/src/heal/manager.rs +++ b/crates/heal/src/heal/manager.rs @@ -220,6 +220,11 @@ struct CompletedHealStatus { result_items: Vec, result_items_truncated: bool, completed_at: SystemTime, + /// Sequence-stamped retained window, archived with the completion so + /// incremental consumers keep their cursor across the transition (HS-06). + seqed_items: Vec<(u64, HealResultItem)>, + next_seq: u64, + min_seq: u64, } #[derive(Debug, Clone)] @@ -240,6 +245,65 @@ pub struct HealTaskReport { pub result_items: Vec, pub result_items_truncated: bool, pub progress: Option, + /// Cursor for incremental consumption: sequence number of the next item + /// to be produced. `0` on reports from sources without sequencing. + pub next_seq: u64, + /// Oldest sequence still retained (`0` together with `next_seq` when + /// sequencing is unavailable). + pub min_seq: u64, +} + +/// Report from a live task, honoring the client's incremental cursor. +async fn active_task_report(task: &HealTask, since: Option) -> HealTaskReport { + let window = task.get_result_items_since(since).await; + HealTaskReport { + status: task.get_status().await, + result_items: window.items, + // The legacy flag stays set once anything was evicted; a lagging + // incremental cursor additionally marks this response truncated so + // the client knows to restart from `min_seq`. + result_items_truncated: task.result_items_truncated() || window.lagged, + progress: Some(task.get_progress().await), + next_seq: window.next_seq, + min_seq: window.min_seq, + } +} + +fn empty_task_report(status: HealTaskStatus) -> HealTaskReport { + HealTaskReport { + status, + result_items: Vec::new(), + result_items_truncated: false, + progress: None, + next_seq: 0, + min_seq: 0, + } +} + +fn completed_task_report(completed: &CompletedHealStatus, since: Option) -> HealTaskReport { + let mut lagged = false; + let result_items = match since { + None => completed.result_items.clone(), + Some(cursor) => { + if cursor + 1 < completed.min_seq { + lagged = true; + } + completed + .seqed_items + .iter() + .filter(|(seq, _)| *seq > cursor) + .map(|(_, item)| item.clone()) + .collect() + } + }; + HealTaskReport { + status: completed.status.clone(), + result_items, + result_items_truncated: completed.result_items_truncated || lagged, + progress: None, + next_seq: completed.next_seq, + min_seq: completed.min_seq, + } } #[derive(Debug, Clone, Copy, Default, PartialEq, Eq, serde::Deserialize, serde::Serialize)] @@ -531,6 +595,11 @@ impl PriorityHealQueue { self.dedup_keys.contains_key(&key) } + /// Iterate queued requests (used by the admin overlap check). + fn requests(&self) -> impl Iterator { + self.heap.iter().map(|item| &item.request) + } + fn contains_request_id(&self, request_id: &str) -> bool { self.heap.iter().any(|item| item.request.id == request_id) } @@ -689,6 +758,80 @@ fn recoverable_heal_retry_delay(retry_attempt: u32) -> Duration { } /// Heal config +/// HS-06 admin overlap policy. +#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)] +pub enum HealOverlapPolicy { + /// Default: overlapping admin starts merge into the existing task + /// (today's dedup semantics). + #[default] + Merge, + /// Return a typed already-running / overlapping-paths rejection like + /// madmin's ErrHealAlreadyRunning / ErrHealOverlappingPaths. + MinioError, +} + +/// Path view of a heal type for overlap comparison: a bucket plus a +/// prefix/object path inside it (`None` bucket = cluster-wide, overlaps +/// everything). +fn heal_type_path_view(heal_type: &HealType) -> (Option<&str>, &str) { + match heal_type { + HealType::Cluster => (None, ""), + HealType::Bucket { bucket } => (Some(bucket), ""), + HealType::Prefix { bucket, prefix } => (Some(bucket), prefix), + HealType::Object { bucket, object, .. } + | HealType::Metadata { bucket, object } + | HealType::ECDecode { bucket, object, .. } => (Some(bucket), object), + // MRF/MetaPath heal keys on a meta path; treat the whole set of + // buckets as one namespace so it only overlaps itself exactly. + HealType::MRF { meta_path } => (Some("\u{0}mrf"), meta_path), + // Erasure-set heal: the set id is the overlap dimension. + HealType::ErasureSet { set_disk_id, .. } => (Some("\u{0}set"), set_disk_id), + } +} + +/// How two heal paths relate for the admin overlap check (HS-06). +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +enum OverlapVerdict { + /// Distinct targets: no conflict. + Disjoint, + /// Same target: an identical heal is already in flight. + SameTarget, + /// One target contains the other. + Overlapping, +} + +fn prefix_paths_overlap(a: &str, b: &str) -> OverlapVerdict { + if a == b { + return OverlapVerdict::SameTarget; + } + if a.is_empty() || b.is_empty() || a.starts_with(b) || b.starts_with(a) { + return OverlapVerdict::Overlapping; + } + OverlapVerdict::Disjoint +} + +fn heal_types_overlap(left: &HealType, right: &HealType) -> OverlapVerdict { + let (left_bucket, left_path) = heal_type_path_view(left); + let (right_bucket, right_path) = heal_type_path_view(right); + match (left_bucket, right_bucket) { + // Cluster-wide overlaps everything (but an exact cluster match is + // SameTarget). + (None, _) | (_, None) => { + if matches!(left, HealType::Cluster) && matches!(right, HealType::Cluster) { + OverlapVerdict::SameTarget + } else { + OverlapVerdict::Overlapping + } + } + (Some(lb), Some(rb)) => { + if lb != rb { + return OverlapVerdict::Disjoint; + } + prefix_paths_overlap(left_path, right_path) + } + } +} + #[derive(Debug, Clone)] pub struct HealConfig { /// Whether to enable auto heal @@ -709,6 +852,9 @@ pub struct HealConfig { pub low_priority_drop_when_full: bool, /// Whether notify-driven scheduler wakeups are enabled. pub event_driven_scheduler_enable: bool, + /// How admin heal starts behave on path overlap (HS-06): merge into the + /// existing task (default) or return a typed already-running rejection. + pub overlap_policy: HealOverlapPolicy, /// Whether per-set bulkhead scheduling is enabled. pub set_bulkhead_enable: bool, /// Whether erasure-set page parallelism is enabled. @@ -757,6 +903,14 @@ impl Default for HealConfig { rustfs_config::ENV_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE, rustfs_config::DEFAULT_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE, ); + let overlap_policy = + match rustfs_utils::get_env_str(rustfs_config::ENV_HEAL_OVERLAP_POLICY, rustfs_config::DEFAULT_HEAL_OVERLAP_POLICY) + .to_lowercase() + .as_str() + { + "minio_error" => HealOverlapPolicy::MinioError, + _ => HealOverlapPolicy::Merge, + }; let set_bulkhead_enable = rustfs_utils::get_env_bool( rustfs_config::ENV_HEAL_SET_BULKHEAD_ENABLE, rustfs_config::DEFAULT_HEAL_SET_BULKHEAD_ENABLE, @@ -793,6 +947,7 @@ impl Default for HealConfig { low_priority_merge_enable, low_priority_drop_when_full, event_driven_scheduler_enable, + overlap_policy, set_bulkhead_enable, page_parallel_enable, mainline_throttle_enable, @@ -1759,6 +1914,50 @@ impl HealManager { request: HealRequest, preserve_alias: bool, ) -> Result { + // HS-06 forceStart semantics (admin only): MinIO stops the old task + // first and then starts the new one. Cancel any active admin task + // overlapping this request's path before entering admission, so the + // fresh task is never merged into the one being replaced. + if request.source == HealRequestSource::Admin && request.force_start { + let overlapping: Vec = { + let active_heals = self.active_heals.lock().await; + active_heals + .iter() + .filter(|(task_id, task)| { + task.source == HealRequestSource::Admin + && heal_types_overlap(&request.heal_type, &task.heal_type) != OverlapVerdict::Disjoint + && *task_id != &request.id + }) + .map(|(task_id, _)| task_id.clone()) + .collect() + }; + for task_id in overlapping { + match self.cancel_task(&task_id).await { + Ok(_) => info!( + target: "rustfs::heal::manager", + event = EVENT_HEAL_QUEUE_ADMISSION, + component = LOG_COMPONENT_HEAL, + subsystem = LOG_SUBSYSTEM_MANAGER, + request_id = %request.id, + cancelled_task_id = %task_id, + result = "force_start_cancelled_overlap", + "Admin forceStart cancelled an overlapping heal task" + ), + Err(err) => warn!( + target: "rustfs::heal::manager", + event = EVENT_HEAL_QUEUE_ADMISSION, + component = LOG_COMPONENT_HEAL, + subsystem = LOG_SUBSYSTEM_MANAGER, + request_id = %request.id, + cancelled_task_id = %task_id, + error = %err, + result = "force_start_cancel_failed", + "Admin forceStart failed to cancel an overlapping heal task" + ), + } + } + } + let config = self.config.read().await; let dedup_key = PriorityHealQueue::make_dedup_key(&request); @@ -1781,7 +1980,15 @@ impl HealManager { .or_else(|| retrying_heal_for_dedup_key(&retrying_heals, &dedup_key).map(|(task_id, _)| (task_id, "retrying"))) }); if let Some((merged_task_id, duplicate_state)) = duplicate.flatten() { - let admission = Self::duplicate_admission_for_request(&request, &config); + // HS-06: under the minio_error overlap policy an exact duplicate + // admin start reports the typed AlreadyRunning rejection instead + // of the silent merge (MinIO's ErrHealAlreadyRunning). + let admission = + if request.source == HealRequestSource::Admin && config.overlap_policy == HealOverlapPolicy::MinioError { + HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning) + } else { + Self::duplicate_admission_for_request(&request, &config) + }; drop(retrying_heals); drop(queue); drop(active_heals); @@ -1827,6 +2034,62 @@ impl HealManager { }); } + // HS-06 typed overlap rejection (admin only, minio_error policy): + // paths containing or contained by an active/queued task reject with + // AlreadyRunning / OverlappingPaths instead of merging. Exact + // duplicates already merged above; scanner/autoheal/read-repair + // sources never take this path. + if request.source == HealRequestSource::Admin && config.overlap_policy == HealOverlapPolicy::MinioError { + let mut rejection = None; + for (task_id, task) in active_heals.iter() { + match heal_types_overlap(&request.heal_type, &task.heal_type) { + OverlapVerdict::SameTarget => { + rejection = Some((HealAdmissionDropReason::AlreadyRunning, task_id.clone())); + break; + } + OverlapVerdict::Overlapping => { + rejection = Some((HealAdmissionDropReason::OverlappingPaths, task_id.clone())); + } + OverlapVerdict::Disjoint => {} + } + } + if rejection.is_none() { + for queued in queue.requests() { + match heal_types_overlap(&request.heal_type, &queued.heal_type) { + OverlapVerdict::SameTarget => { + rejection = Some((HealAdmissionDropReason::AlreadyRunning, queued.id.clone())); + break; + } + OverlapVerdict::Overlapping => { + rejection = Some((HealAdmissionDropReason::OverlappingPaths, queued.id.clone())); + } + OverlapVerdict::Disjoint => {} + } + } + } + if let Some((reason, overlap_task_id)) = rejection { + drop(retrying_heals); + drop(queue); + drop(active_heals); + Self::record_admission_metric(request.source, HealAdmissionResult::Dropped(reason), "overlap_rejected"); + warn!( + target: "rustfs::heal::manager", + event = EVENT_HEAL_QUEUE_ADMISSION, + component = LOG_COMPONENT_HEAL, + subsystem = LOG_SUBSYSTEM_MANAGER, + request_id = %request.id, + overlap_task_id = %overlap_task_id, + reason = reason.as_str(), + result = "overlap_rejected", + "Admin heal start rejected by overlap policy" + ); + return Ok(HealAdmissionReceipt { + result: HealAdmissionResult::Dropped(reason), + task_id: overlap_task_id, + }); + } + } + let mut task_id = request.id.clone(); let admission = Self::admit_request_to_queue(&mut queue, request, &config, "submit"); if admission == HealAdmissionResult::Merged @@ -1899,28 +2162,25 @@ impl HealManager { } pub async fn get_task_report(&self, task_id: &str) -> Result { + self.get_task_report_since(task_id, None).await + } + + /// Incremental variant of [`Self::get_task_report`] (HS-06): `since` is + /// the client's last seen sequence number; `None` keeps the legacy + /// full-snapshot semantics. + pub async fn get_task_report_since(&self, task_id: &str, since: Option) -> Result { let canonical_task_id = self.canonical_task_id(task_id).await; { let active_heals = self.active_heals.lock().await; if let Some(task) = active_heals.get(&canonical_task_id) { - return Ok(HealTaskReport { - status: task.get_status().await, - result_items: task.get_result_items().await, - result_items_truncated: task.result_items_truncated(), - progress: Some(task.get_progress().await), - }); + return Ok(active_task_report(task, since).await); } } { let retrying_heals = self.retrying_heals.lock().await; if let Some(retrying) = retrying_heals.get(&canonical_task_id) { - return Ok(HealTaskReport { - status: retrying.status(), - result_items: Vec::new(), - result_items_truncated: false, - progress: None, - }); + return Ok(empty_task_report(retrying.status())); } } @@ -1930,36 +2190,21 @@ impl HealManager { if let Some(completed) = completed_heals.get(&canonical_task_id) && completed_status_is_retrying(&completed.status) { - return Ok(HealTaskReport { - status: completed.status.clone(), - result_items: completed.result_items.clone(), - result_items_truncated: completed.result_items_truncated, - progress: None, - }); + return Ok(completed_task_report(completed, since)); } } { let queue = self.heal_queue.lock().await; if queue.contains_request_id(&canonical_task_id) { - return Ok(HealTaskReport { - status: HealTaskStatus::Pending, - result_items: Vec::new(), - result_items_truncated: false, - progress: None, - }); + return Ok(empty_task_report(HealTaskStatus::Pending)); } } let mut completed_heals = self.completed_heals.lock().await; prune_completed_heal_statuses(&mut completed_heals); if let Some(completed) = completed_heals.get(&canonical_task_id) { - return Ok(HealTaskReport { - status: completed.status.clone(), - result_items: completed.result_items.clone(), - result_items_truncated: completed.result_items_truncated, - progress: None, - }); + return Ok(completed_task_report(completed, since)); } Err(Error::TaskNotFound { @@ -1968,18 +2213,23 @@ impl HealManager { } pub async fn get_task_report_for_path(&self, heal_path: &str, task_id: &str) -> Result { + self.get_task_report_for_path_since(heal_path, task_id, None).await + } + + /// Incremental variant of [`Self::get_task_report_for_path`] (HS-06). + pub async fn get_task_report_for_path_since( + &self, + heal_path: &str, + task_id: &str, + since: Option, + ) -> Result { let canonical_task_id = self.canonical_task_id(task_id).await; { let active_heals = self.active_heals.lock().await; if let Some(task) = active_heals.get(&canonical_task_id) && heal_type_matches_path(&task.heal_type, heal_path) { - return Ok(HealTaskReport { - status: task.get_status().await, - result_items: task.get_result_items().await, - result_items_truncated: task.result_items_truncated(), - progress: Some(task.get_progress().await), - }); + return Ok(active_task_report(task, since).await); } } @@ -1988,12 +2238,7 @@ impl HealManager { if let Some(retrying) = retrying_heals.get(&canonical_task_id) && heal_type_matches_path(&retrying.request.heal_type, heal_path) { - return Ok(HealTaskReport { - status: retrying.status(), - result_items: Vec::new(), - result_items_truncated: false, - progress: None, - }); + return Ok(empty_task_report(retrying.status())); } } @@ -2004,24 +2249,14 @@ impl HealManager { && heal_type_matches_path(&completed.heal_type, heal_path) && completed_status_is_retrying(&completed.status) { - return Ok(HealTaskReport { - status: completed.status.clone(), - result_items: completed.result_items.clone(), - result_items_truncated: completed.result_items_truncated, - progress: None, - }); + return Ok(completed_task_report(completed, since)); } } { let queue = self.heal_queue.lock().await; if queue.contains_request_id_matching_path(&canonical_task_id, heal_path) { - return Ok(HealTaskReport { - status: HealTaskStatus::Pending, - result_items: Vec::new(), - result_items_truncated: false, - progress: None, - }); + return Ok(empty_task_report(HealTaskStatus::Pending)); } } @@ -2031,12 +2266,7 @@ impl HealManager { if let Some(completed) = completed_heals.get(&canonical_task_id) && heal_type_matches_path(&completed.heal_type, heal_path) { - return Ok(HealTaskReport { - status: completed.status.clone(), - result_items: completed.result_items.clone(), - result_items_truncated: completed.result_items_truncated, - progress: None, - }); + return Ok(completed_task_report(completed, since)); } } @@ -3231,12 +3461,16 @@ impl HealManager { completed_task.get_status().await }; let completed_progress = completed_task.get_progress().await; + let final_window = completed_task.get_result_items_since(None).await; let completed_status_entry = CompletedHealStatus { heal_type: completed_task.heal_type.clone(), status: completed_status.clone(), - result_items: completed_task.get_result_items().await, + result_items: final_window.items.clone(), result_items_truncated: completed_task.result_items_truncated(), completed_at: SystemTime::now(), + seqed_items: completed_task.get_seqed_result_items().await, + next_seq: final_window.next_seq, + min_seq: final_window.min_seq, }; let mut completed_heals_guard = completed_heals_clone.lock().await; prune_completed_heal_statuses(&mut completed_heals_guard); @@ -5008,6 +5242,9 @@ mod tests { }, result_items: Vec::new(), result_items_truncated: false, + seqed_items: Vec::new(), + next_seq: 0, + min_seq: 0, completed_at: SystemTime::now(), }, ); @@ -5289,6 +5526,136 @@ mod tests { assert_eq!(snapshot.queued_by_source.internal, 0); } + // HS-06 (backlog#1870): overlap policy + forceStart semantics. + fn manager_with_policy(policy: HealOverlapPolicy) -> HealManager { + let storage: Arc = Arc::new(MockStorage); + HealManager::new( + storage, + Some(HealConfig { + overlap_policy: policy, + ..Default::default() + }), + ) + } + + fn admin_prefix_request(bucket: &str, prefix: &str) -> HealRequest { + let mut request = HealRequest::new( + HealType::Prefix { + bucket: bucket.to_string(), + prefix: prefix.to_string(), + }, + HealOptions::default(), + HealPriority::Normal, + ); + request.source = HealRequestSource::Admin; + request + } + + async fn insert_active_task(manager: &HealManager, request: HealRequest) -> String { + let task = Arc::new(HealTask::from_request(request, manager.storage.clone())); + let task_id = task.id.clone(); + manager.active_heals.lock().await.insert(task_id.clone(), task); + task_id + } + + #[tokio::test] + async fn overlap_policy_minio_error_rejects_same_and_containing_paths() { + let manager = manager_with_policy(HealOverlapPolicy::MinioError); + insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await; + + // Same target: typed AlreadyRunning. + let same = manager + .submit_heal_request(admin_prefix_request("bucket-a", "logs/")) + .await + .expect("admission must decide"); + assert_eq!( + same, + HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning), + "an identical target must reject with already-running" + ); + + // Contained path: typed OverlappingPaths. + let nested = manager + .submit_heal_request(admin_prefix_request("bucket-a", "logs/app/")) + .await + .expect("admission must decide"); + assert_eq!( + nested, + HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths), + "a path inside the active task's path must reject with overlapping-paths" + ); + + // Containing path (bucket-wide vs nested active): also overlapping. + let wide = manager + .submit_heal_request(admin_prefix_request("bucket-a", "")) + .await + .expect("admission must decide"); + assert_eq!( + wide, + HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths), + "a bucket-wide start overlapping a nested active heal must reject" + ); + + // Disjoint bucket: unaffected. + let disjoint = manager + .submit_heal_request(admin_prefix_request("bucket-b", "logs/")) + .await + .expect("admission must decide"); + assert_eq!(disjoint, HealAdmissionResult::Accepted); + } + + #[tokio::test] + async fn overlap_policy_default_merge_keeps_today_semantics() { + let manager = manager_with_policy(HealOverlapPolicy::Merge); + insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await; + + // Different-dedup-key overlap still merges under the default policy: + // the nested path dedups to its own key but nothing rejects it. + let nested = manager + .submit_heal_request(admin_prefix_request("bucket-a", "logs/app/")) + .await + .expect("admission must decide"); + assert_eq!(nested, HealAdmissionResult::Accepted, "default policy must not reject overlaps"); + + // Non-admin sources never get overlap rejections even under minio_error. + let manager = manager_with_policy(HealOverlapPolicy::MinioError); + insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await; + let mut scanner_request = admin_prefix_request("bucket-a", "logs/app/"); + scanner_request.source = HealRequestSource::Scanner; + let admitted = manager + .submit_heal_request(scanner_request) + .await + .expect("admission must decide"); + assert_eq!(admitted, HealAdmissionResult::Accepted, "scanner sources must never be overlap-rejected"); + } + + #[tokio::test] + async fn admin_force_start_cancels_overlapping_active_task_first() { + let manager = manager_with_policy(HealOverlapPolicy::Merge); + let old_id = insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await; + + let mut replacement = admin_prefix_request("bucket-a", "logs/"); + replacement.force_start = true; + let receipt = manager + .submit_heal_request_with_receipt(replacement) + .await + .expect("force-start submission must decide"); + + assert!(receipt.result.is_admitted(), "the new task must be admitted (Accepted or Merged)"); + let old_task_gone = { + let active_heals = manager.active_heals.lock().await; + !active_heals.contains_key(&old_id) + }; + assert!( + old_task_gone, + "the overlapping admin task must be cancelled (removed from the active table) before the new one starts" + ); + assert!( + matches!(manager.get_task_status(&old_id).await, Err(Error::TaskNotFound { .. })), + "a cancelled task must no longer resolve as an active heal" + ); + } + #[tokio::test] async fn test_operations_snapshot_counts_active_by_source_and_priority() { let storage: Arc = Arc::new(MockStorage); @@ -5591,6 +5958,9 @@ mod tests { status: HealTaskStatus::Completed, result_items: Vec::new(), result_items_truncated: false, + seqed_items: Vec::new(), + next_seq: 0, + min_seq: 0, completed_at: SystemTime::now(), }, ); @@ -5625,6 +5995,9 @@ mod tests { ..Default::default() }], result_items_truncated: true, + seqed_items: Vec::new(), + next_seq: 0, + min_seq: 0, completed_at: SystemTime::now(), }, ); diff --git a/crates/heal/src/heal/task.rs b/crates/heal/src/heal/task.rs index f6472eb65..2c0626d8f 100644 --- a/crates/heal/src/heal/task.rs +++ b/crates/heal/src/heal/task.rs @@ -32,7 +32,7 @@ use std::{ future::Future, sync::{ Arc, - atomic::{AtomicBool, Ordering}, + atomic::{AtomicBool, AtomicU64, Ordering}, }, time::{Duration, Instant, SystemTime}, }; @@ -351,6 +351,20 @@ impl HealRequest { } /// Heal task +/// Incremental view over a task's retained result items (HS-06). +/// +/// `next_seq` is the cursor a client should pass on its next poll; `min_seq` +/// is the oldest sequence still retained; `lagged` means the client's cursor +/// fell behind `min_seq` and items were skipped — the client should restart +/// from `min_seq`. +#[derive(Debug, Clone)] +pub struct HealResultWindow { + pub items: Vec, + pub next_seq: u64, + pub min_seq: u64, + pub lagged: bool, +} + pub struct HealTask { /// Task ID pub id: String, @@ -373,8 +387,16 @@ pub struct HealTask { pub status: Arc>, /// Progress tracking pub progress: Arc>, - /// Result items collected from storage heal calls. - pub result_items: Arc>>, + /// Result items collected from storage heal calls, each stamped with a + /// monotonically increasing sequence number for incremental consumption + /// (the client passes the last seen seq back and receives only newer + /// items; see `get_result_items_since`). + pub result_items: Arc>>, + /// Next sequence number to assign; starts at 1. + next_item_seq: Arc, + /// Sequence number of the oldest item still inside the retention window; + /// equals `next_item_seq` while the window is empty. + min_available_seq: Arc, result_items_truncated: Arc, batch_failure: Arc>>, batch_failure_recorded: Arc, @@ -426,6 +448,8 @@ impl HealTask { status: Arc::new(RwLock::new(HealTaskStatus::Pending)), progress: Arc::new(RwLock::new(HealProgress::new())), result_items: Arc::new(RwLock::new(Vec::new())), + next_item_seq: Arc::new(AtomicU64::new(1)), + min_available_seq: Arc::new(AtomicU64::new(1)), result_items_truncated: Arc::new(AtomicBool::new(false)), batch_failure: Arc::new(RwLock::new(None)), batch_failure_recorded: Arc::new(AtomicBool::new(false)), @@ -911,18 +935,63 @@ impl HealTask { } pub async fn get_result_items(&self) -> Vec { + self.result_items.read().await.iter().map(|(_, item)| item.clone()).collect() + } + + /// Sequence-stamped retained window, used when archiving a completed + /// task so incremental cursors survive the transition (HS-06). + pub async fn get_seqed_result_items(&self) -> Vec<(u64, HealResultItem)> { self.result_items.read().await.clone() } + /// Incremental result window (HS-06): `since = None` returns the full + /// retained window (legacy snapshot semantics); `since = Some(seq)` + /// returns only items stamped with a sequence greater than `seq`. + /// `lagged` warns that the caller's cursor fell behind the window start + /// and items were skipped (the response carries `min_seq` as the catch-up + /// cursor). + pub async fn get_result_items_since(&self, since: Option) -> HealResultWindow { + let result_items = self.result_items.read().await; + let next_seq = self.next_item_seq.load(Ordering::Relaxed); + let min_seq = self.min_available_seq.load(Ordering::Relaxed); + let mut lagged = false; + let items = match since { + None => result_items.iter().map(|(_, item)| item.clone()).collect::>(), + Some(cursor) => { + if cursor + 1 < min_seq { + lagged = true; + } + result_items + .iter() + .filter(|(seq, _)| *seq > cursor) + .map(|(_, item)| item.clone()) + .collect::>() + } + }; + HealResultWindow { + items, + next_seq, + min_seq, + lagged, + } + } + pub fn result_items_truncated(&self) -> bool { self.result_items_truncated.load(Ordering::Relaxed) } async fn record_result_item(&self, result: HealResultItem) { + let seq = self.next_item_seq.fetch_add(1, Ordering::Relaxed); let mut result_items = self.result_items.write().await; if result_items.len() < MAX_RETAINED_HEAL_RESULT_ITEMS { - result_items.push(result); + result_items.push((seq, result)); } else { + // Slide the window: the oldest item leaves and the cursor for the + // oldest still-available item moves forward with it. + result_items.remove(0); + self.min_available_seq + .store(result_items.first().map_or(seq, |(oldest, _)| *oldest), Ordering::Relaxed); + result_items.push((seq, result)); self.result_items_truncated.store(true, Ordering::Relaxed); } } @@ -3880,6 +3949,69 @@ mod tests { assert!(task.result_items_truncated()); } + // HS-06 (backlog#1870): incremental result windows. + #[tokio::test] + async fn result_items_seq_is_monotonic_and_incremental_slices_work() { + let storage = Arc::new(MockStorage::default()); + let task = HealTask::from_request(HealRequest::bucket("bucket-a".to_string()), storage); + + for round in 0..5u64 { + let item = HealResultItem { + object_size: round as usize, + ..Default::default() + }; + task.record_result_item(item).await; + } + + let full = task.get_result_items_since(None).await; + assert_eq!(full.items.len(), 5, "None keeps the full-snapshot semantics"); + assert_eq!(full.next_seq, 6, "next_seq is one past the last assigned"); + assert_eq!(full.min_seq, 1, "nothing was evicted yet"); + assert!(!full.lagged); + + // Incremental: only items newer than the cursor. + let incremental = task.get_result_items_since(Some(3)).await; + assert_eq!( + incremental.items.iter().map(|item| item.object_size).collect::>(), + vec![3, 4], + "only sequences greater than the cursor are returned" + ); + assert_eq!(incremental.next_seq, 6); + + // A cursor at the head is not lagging. + assert!(!task.get_result_items_since(Some(0)).await.lagged); + } + + #[tokio::test] + async fn result_items_window_slide_moves_min_seq_and_flags_lagging_cursors() { + let storage = Arc::new(MockStorage::default()); + let task = HealTask::from_request(HealRequest::bucket("bucket-a".to_string()), storage); + + // Fill the window completely, then push two more items: seq 1 and 2 + // are evicted by the slide. + for _ in 0..(MAX_RETAINED_HEAL_RESULT_ITEMS + 2) { + task.record_result_item(HealResultItem::default()).await; + } + + let full = task.get_result_items_since(None).await; + assert_eq!(full.items.len(), MAX_RETAINED_HEAL_RESULT_ITEMS); + assert_eq!(full.min_seq, 3, "each evicted head item moved the oldest-available cursor"); + assert!(task.result_items_truncated()); + + // A client still polling from before the eviction is lagging. + let lagging = task.get_result_items_since(Some(0)).await; + assert!(lagging.lagged, "a cursor behind min_seq must be flagged"); + assert_eq!(lagging.min_seq, 3, "the response tells the client where to restart"); + + // A cursor inside the window is fine. + assert!(!task.get_result_items_since(Some(3)).await.lagged); + + // The lagging client restarts from min_seq and gets the full window. + let catch_up = task.get_result_items_since(Some(3)).await; + assert_eq!(catch_up.items.len(), MAX_RETAINED_HEAL_RESULT_ITEMS - 1); + assert!(!catch_up.lagged); + } + #[tokio::test] async fn test_recursive_bucket_heal_skips_object_dir_candidates() { let storage = Arc::new(MockStorage { diff --git a/crates/protos/src/heal_control.rs b/crates/protos/src/heal_control.rs index df89126d2..515c0db75 100644 --- a/crates/protos/src/heal_control.rs +++ b/crates/protos/src/heal_control.rs @@ -177,16 +177,38 @@ impl StartCommand { #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)] #[serde(tag = "action", rename_all = "snake_case", deny_unknown_fields)] pub enum Command { - Start { request: StartCommand }, - Query { heal_path: String, client_token: String }, - Cancel { heal_path: String, client_token: String }, + Start { + request: StartCommand, + }, + Query { + heal_path: String, + client_token: String, + /// Incremental result cursor (HS-06): only items with a sequence + /// greater than this are returned. Absent = legacy full snapshot. + /// Optional + defaulted so older peers stay wire-compatible. + #[serde(default, skip_serializing_if = "Option::is_none")] + since_seq: Option, + }, + Cancel { + heal_path: String, + client_token: String, + }, } #[derive(Debug)] pub enum ExecutableCommand { - Start { request: HealChannelRequest }, - Query { heal_path: String, client_token: String }, - Cancel { heal_path: String, client_token: String }, + Start { + request: HealChannelRequest, + }, + Query { + heal_path: String, + client_token: String, + since_seq: Option, + }, + Cancel { + heal_path: String, + client_token: String, + }, } #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)] @@ -227,8 +249,22 @@ impl Envelope { ) } - pub fn query(request_id: String, metadata: RequestMetadata, heal_path: String, client_token: String) -> Result { - Self::new(request_id, metadata, Command::Query { heal_path, client_token }) + pub fn query( + request_id: String, + metadata: RequestMetadata, + heal_path: String, + client_token: String, + since_seq: Option, + ) -> Result { + Self::new( + request_id, + metadata, + Command::Query { + heal_path, + client_token, + since_seq, + }, + ) } pub fn cancel( @@ -286,7 +322,15 @@ impl Envelope { Command::Start { request } => ExecutableCommand::Start { request: request.into_channel_request(self.request_id.clone())?, }, - Command::Query { heal_path, client_token } => ExecutableCommand::Query { heal_path, client_token }, + Command::Query { + heal_path, + client_token, + since_seq, + } => ExecutableCommand::Query { + heal_path, + client_token, + since_seq, + }, Command::Cancel { heal_path, client_token } => ExecutableCommand::Cancel { heal_path, client_token }, }; Ok((self.request_id, self.coordinator_epoch, command)) @@ -305,6 +349,11 @@ pub enum Admission { Full, DroppedQueueFull, DroppedPolicy, + /// HS-06: admin start rejected because the same target is already being + /// healed (RUSTFS_HEAL_OVERLAP_POLICY=minio_error only). + DroppedAlreadyRunning, + /// HS-06: admin start rejected because its path overlaps an active heal. + DroppedOverlappingPaths, } impl From for Admission { @@ -315,6 +364,8 @@ impl From for Admission { HealAdmissionResult::Full => Self::Full, HealAdmissionResult::Dropped(HealAdmissionDropReason::QueueFull) => Self::DroppedQueueFull, HealAdmissionResult::Dropped(HealAdmissionDropReason::PolicyDropped) => Self::DroppedPolicy, + HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning) => Self::DroppedAlreadyRunning, + HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths) => Self::DroppedOverlappingPaths, } } } @@ -331,6 +382,8 @@ impl Admission { Self::Full => HealAdmissionResult::Full, Self::DroppedQueueFull => HealAdmissionResult::Dropped(HealAdmissionDropReason::QueueFull), Self::DroppedPolicy => HealAdmissionResult::Dropped(HealAdmissionDropReason::PolicyDropped), + Self::DroppedAlreadyRunning => HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning), + Self::DroppedOverlappingPaths => HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths), } } } @@ -592,6 +645,7 @@ mod tests { metadata(2, 7), "bucket/prefix".to_string(), "token".to_string(), + None, ) .unwrap(); let cancel = Envelope::cancel( @@ -667,6 +721,7 @@ mod tests { RequestMetadata::new([0x11; 16], 1_700_000_000_000, 1_700_000_030_000, 9), "bucket/prefix".to_string(), "client-token".to_string(), + None, ) .unwrap(); let cancel = Envelope::cancel( @@ -749,7 +804,7 @@ mod tests { assert!(Envelope::start(test_request(request_id.clone()), metadata(0, 7)).is_err()); assert!(Envelope::start(test_request(request_id.clone()), metadata(1, 0)).is_err()); assert!(Envelope::start(test_request(request_id.clone()), RequestMetadata::new([1; 16], 1_000, 31_001, 7),).is_err()); - assert!(Envelope::query(request_id.clone(), metadata(1, 7), String::new(), String::new()).is_err()); + assert!(Envelope::query(request_id.clone(), metadata(1, 7), String::new(), String::new(), None).is_err()); assert!(Envelope::cancel(request_id.clone(), metadata(1, 7), String::new(), String::new()).is_ok()); let mut noncanonical_request = test_request(request_id.to_uppercase()); @@ -782,6 +837,7 @@ mod tests { metadata(1, 7), "x".repeat(ENVELOPE_MAX_SIZE), "token".to_string(), + None, ) .unwrap(); let error = super::encode_envelope(&oversized).unwrap_err(); diff --git a/crates/scanner/src/scanner_folder.rs b/crates/scanner/src/scanner_folder.rs index 030c55fc9..5766385e8 100644 --- a/crates/scanner/src/scanner_folder.rs +++ b/crates/scanner/src/scanner_folder.rs @@ -1770,6 +1770,13 @@ impl FolderScanner { HealAdmissionResult::Dropped(HealAdmissionDropReason::PolicyDropped) => { self.clear_pending_scanner_heal(kind, bucket, object, version_id); } + // Admin-only overlap rejections (HS-06); the scanner never sees + // them, but if it ever does, treat them as terminal like any + // other policy drop rather than endlessly retrying. + HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning) + | HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths) => { + self.clear_pending_scanner_heal(kind, bucket, object, version_id); + } } } diff --git a/rustfs/src/admin/handlers/heal.rs b/rustfs/src/admin/handlers/heal.rs index 52e9fc152..3d98e966a 100644 --- a/rustfs/src/admin/handlers/heal.rs +++ b/rustfs/src/admin/handlers/heal.rs @@ -66,6 +66,9 @@ struct HealInitParams { client_token: String, force_start: bool, force_stop: bool, + /// Incremental result cursor (HS-06): only result items with a sequence + /// greater than this are returned; absent means full snapshot. + since_seq: Option, } fn extract_heal_init_params(body: &Bytes, uri: &Uri, params: Params<'_, '_>) -> S3Result { @@ -98,6 +101,16 @@ fn extract_heal_init_params(body: &Bytes, uri: &Uri, params: Params<'_, '_>) -> } hip.force_stop = parse_heal_query_bool(value.as_ref())?; } + "sinceSeq" => { + if !seen.insert("sinceSeq") { + return Err(s3_error!(InvalidArgument, "duplicate heal query parameter")); + } + hip.since_seq = Some( + value + .parse::() + .map_err(|_| s3_error!(InvalidArgument, "sinceSeq must be a non-negative integer"))?, + ); + } _ => return Err(s3_error!(InvalidArgument, "unknown heal query parameter")), } } @@ -978,7 +991,15 @@ fn reject_heal_admission(result: rustfs_common::heal_channel::HealAdmissionResul result.result_label(), result.reason_label() ), - HealAdmissionResult::Dropped(HealAdmissionDropReason::PolicyDropped) => s3_error!( + // Overlap rejections (HS-06) share this arm: the s3s footprint + // ratchet forbids new s3_error! sites, and the typed reason is + // preserved through reason_label() ("already_running" / + // "overlapping_paths") so madmin-style clients can distinguish. + HealAdmissionResult::Dropped( + HealAdmissionDropReason::PolicyDropped + | HealAdmissionDropReason::AlreadyRunning + | HealAdmissionDropReason::OverlappingPaths, + ) => s3_error!( OperationAborted, "heal request not admitted: admission={}, reason={}", result.result_label(), @@ -1403,6 +1424,7 @@ impl Operation for HealHandler { new_heal_control_metadata(&route)?, heal_path, client_token.clone(), + hip.since_seq, ) .map_err(|err| s3_error!(InternalError, "encode heal control query failed: {err}"))?; let response = submit_cluster_heal_channel_command(context, route, envelope, &request_id, client_token).await?; diff --git a/rustfs/src/storage/rpc/node_service.rs b/rustfs/src/storage/rpc/node_service.rs index a39ca6c3a..352bfbd58 100644 --- a/rustfs/src/storage/rpc/node_service.rs +++ b/rustfs/src/storage/rpc/node_service.rs @@ -571,8 +571,12 @@ async fn execute_heal_control_envelope_with_manager( admission: receipt.result.into(), } } - rustfs_protos::heal_control::ExecutableCommand::Query { heal_path, client_token } => { - let response = timeout(remaining, processor.execute_query_request(heal_path, client_token)) + rustfs_protos::heal_control::ExecutableCommand::Query { + heal_path, + client_token, + since_seq, + } => { + let response = timeout(remaining, processor.execute_query_request_since(heal_path, client_token, since_seq)) .await .map_err(|_| Status::deadline_exceeded("heal control query expired before execution"))? .map_err(|_| Status::internal("heal control query failed"))?; @@ -2517,6 +2521,7 @@ mod tests { metadata(), "bucket/prefix".to_string(), canonical_token.clone(), + None, ) .unwrap(); let query_result = execute_heal_control_envelope_with_manager(query, coordinator_epoch, Some(Arc::clone(&manager))) @@ -2555,6 +2560,7 @@ mod tests { metadata(), "bucket/prefix".to_string(), canonical_token, + None, ) .unwrap(); let stopped_result = execute_heal_control_envelope_with_manager(stopped_query, coordinator_epoch, Some(manager)) diff --git a/scripts/check_s3s_footprint.sh b/scripts/check_s3s_footprint.sh index 56fb60d35..1ebd59f5b 100755 --- a/scripts/check_s3s_footprint.sh +++ b/scripts/check_s3s_footprint.sh @@ -25,7 +25,7 @@ cd "$(dirname "$0")/.." # Baselines verified on 2026-08-11. Lower-only; see header. # Excludes crates/e2e_test/ — test infrastructure legitimately uses s3s # to verify S3 behavior and does not widen the production s3s surface. -S3S_IMPORT_FILES_BASELINE=213 +S3S_IMPORT_FILES_BASELINE=211 S3_ERROR_LINES_BASELINE=1620 S3S_PATH_PATTERN='(^|[^"[:alnum:]_])s3s::' E2E_TEST_GLOB='--glob=!crates/e2e_test/**' From bd296eff9e69d19ae8895e371f443addd5671ec9 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 16:14:05 +0800 Subject: [PATCH 10/24] chore(io-core): drop eight zero-consumer modules (#6201) --- ARCHITECTURE.md | 8 +- Cargo.lock | 1 - crates/io-core/CHANGELOG.md | 3 + crates/io-core/Cargo.toml | 5 +- crates/io-core/README.md | 138 +-- crates/io-core/README_zh.md | 149 +--- crates/io-core/examples/scheduler_example.rs | 190 ---- crates/io-core/src/bufreader_optimizer.rs | 227 ----- crates/io-core/src/direct_io.rs | 332 ------- crates/io-core/src/io_priority_queue.rs | 381 -------- crates/io-core/src/lib.rs | 73 +- crates/io-core/src/progress.rs | 138 +++ crates/io-core/src/reader.rs | 412 --------- crates/io-core/src/scheduler.rs | 882 ------------------- crates/io-core/src/shared_memory.rs | 320 ------- crates/io-core/src/timeout_wrapper.rs | 501 ----------- crates/io-core/src/writer.rs | 443 ---------- 17 files changed, 195 insertions(+), 4008 deletions(-) delete mode 100644 crates/io-core/examples/scheduler_example.rs delete mode 100644 crates/io-core/src/bufreader_optimizer.rs delete mode 100644 crates/io-core/src/direct_io.rs delete mode 100644 crates/io-core/src/io_priority_queue.rs create mode 100644 crates/io-core/src/progress.rs delete mode 100644 crates/io-core/src/reader.rs delete mode 100644 crates/io-core/src/scheduler.rs delete mode 100644 crates/io-core/src/shared_memory.rs delete mode 100644 crates/io-core/src/timeout_wrapper.rs delete mode 100644 crates/io-core/src/writer.rs diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md index 809d0a3f5..4475c2997 100644 --- a/ARCHITECTURE.md +++ b/ARCHITECTURE.md @@ -31,7 +31,7 @@ HTTP request → storage/ecfs (erasure coding, encryption, checksums) → ecstore (disk pool selection, data distribution) → rio (reader pipeline: encrypt → compress → hash → write) - → io-core (zero-copy I/O, buffer pool, direct I/O) + → io-core (buffer pool, storage profiling, admission control) → local disk / remote disk via RPC ``` @@ -55,7 +55,7 @@ rustfs/ # Workspace root (virtual manifest) ├── crates/ # library crates (authoritative list: Cargo.toml [workspace].members) │ ├── ecstore/ # Erasure-coded storage engine │ ├── rio/ # Reader I/O pipeline (encrypt, compress, hash) -│ ├── io-core/ # Zero-copy I/O, scheduling, buffer pool +│ ├── io-core/ # Buffer pool, storage profiling, admission control │ ├── io-metrics/ # I/O metrics collection │ ├── common/ # Shared runtime state, globals, data usage types │ ├── config/ # Configuration types and parsing @@ -302,7 +302,7 @@ The binary (`main.rs`) boots in this order: │ │ │ ┌─────▼──────┐ ┌──────▼──────┐ ┌──────▼──────┐ │ ecstore │ │ rio │ │ io-core │ - │ (core) │ │ (readers) │ │ (zero-copy) │ + │ (core) │ │ (readers) │ │ (buffers) │ └─────┬──────┘ └─────────────┘ └─────────────┘ │ ┌─────┬──┼──┬─────┬──────┐ @@ -314,7 +314,7 @@ The binary (`main.rs`) boots in this order: - **"Where does S3 PutObject go?"** `server/` routes → `app/object_usecase` validates → `storage/ecfs` encodes → - `ecstore` distributes → `rio` encrypts/compresses → `io-core` writes + `ecstore` distributes → `rio` encrypts/compresses → `io-core` supplies buffers - **"Where are bucket policies enforced?"** `app/bucket_usecase` calls into `crates/policy/` diff --git a/Cargo.lock b/Cargo.lock index 83749d574..0fb09e78e 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -9613,7 +9613,6 @@ version = "1.0.0-rc.2" dependencies = [ "bytes", "hotpath", - "memmap2", "rustfs-io-metrics", "thiserror 2.0.20", "tokio", diff --git a/crates/io-core/CHANGELOG.md b/crates/io-core/CHANGELOG.md index 2ed1235c4..b60a494a1 100644 --- a/crates/io-core/CHANGELOG.md +++ b/crates/io-core/CHANGELOG.md @@ -9,6 +9,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0 ### Removed +#### rustfs-io-core +- **Zero-consumer modules** (added in 0.0.5): `reader`, `writer`, `bufreader_optimizer`, `shared_memory`, `direct_io`, `timeout_wrapper`, `io_priority_queue`, and `scheduler` had no caller in the workspace and were removed (rustfs/backlog#1824). The scheduling algorithm and the request timeout wrapper that RustFS actually runs live in `rustfs/src/storage/`; this crate keeps the config shapes they project into. `OperationProgress` moved to the new `progress` module and is still exported as `rustfs_io_core::OperationProgress`. + #### rustfs-io-metrics - **Unified configuration** (added in 0.0.5): the zero-consumer `IoConfig`, `CacheSettings`, `IoSchedulerSettings`, `BackpressureSettings`, `TimeoutSettings`, `DeadlockDetectionSettings` types and their `DEFAULT_*` constants were removed (rustfs/rustfs#6008); rustfs-io-core's `IoSchedulerConfig`/`BackpressureConfig` remain the canonical configuration types. diff --git a/crates/io-core/Cargo.toml b/crates/io-core/Cargo.toml index 9c309633b..f3118c19a 100644 --- a/crates/io-core/Cargo.toml +++ b/crates/io-core/Cargo.toml @@ -20,8 +20,8 @@ license.workspace = true repository.workspace = true rust-version.workspace = true homepage.workspace = true -description = "Buffered I/O reader and writer implementations for RustFS (mmap-then-copy, aligned pread)" -keywords = ["io", "reader", "writer", "rustfs", "mmap"] +description = "Shared I/O primitives for RustFS (buffer pool, storage profiling, backpressure, deadlock detection)" +keywords = ["io", "buffer", "pool", "rustfs", "backpressure"] categories = ["development-tools", "filesystem"] [lints] @@ -38,7 +38,6 @@ hotpath.workspace = true bytes = { workspace = true, features = ["serde"] } thiserror = { workspace = true } tokio = { workspace = true, features = ["io-util", "fs", "sync", "rt-multi-thread"] } -memmap2 = { workspace = true } rustfs-io-metrics = { workspace = true } tracing = { workspace = true } diff --git a/crates/io-core/README.md b/crates/io-core/README.md index ee5f7dd50..a372dab1b 100644 --- a/crates/io-core/README.md +++ b/crates/io-core/README.md @@ -23,67 +23,20 @@ ## Overview -**rustfs-io-core** is the core I/O scheduling module for [RustFS](https://rustfs.com), a distributed object storage system. It provides: +**rustfs-io-core** holds the shared I/O primitives for [RustFS](https://rustfs.com), a distributed object storage system. It provides: -- **I/O Scheduler**: Adaptive buffer size calculation and load management -- **Priority Queue**: Request priority scheduling with starvation prevention +- **Buffer Pool**: Tiered `BytesPool` for buffer reuse +- **Storage Profiling**: Storage-media and access-pattern model (`io_profile`) +- **Scheduler Configuration**: The `IoSchedulerConfig` / `IoPriorityQueueConfig` shapes the storage layer projects into - **Backpressure Control**: System overload protection with graceful degradation - **Deadlock Detection**: Wait-for graph based deadlock detection algorithm - **Lock Optimizer**: Adaptive spin lock optimization -- **Timeout Wrapper**: Dynamic timeout calculation and operation progress tracking +- **Progress Tracking**: Byte progress and staleness for long-running operations + +The scheduling algorithm itself lives in `rustfs/src/storage/concurrency/io_schedule.rs`; this crate carries the configuration shapes it projects into, not a second implementation. ## Features -### I/O Scheduler - -Adaptive I/O scheduling with dynamic buffer size calculation based on file size, access pattern, and system load: - -```rust -use rustfs_io_core::{IoScheduler, IoSchedulerConfig, IoLoadLevel}; -use rustfs_io_core::io_profile::{StorageMedia, AccessPattern}; - -// Create scheduler -let config = IoSchedulerConfig { - max_concurrent_reads: 64, - base_buffer_size: 64 * 1024, // 64 KB - max_buffer_size: 1024 * 1024, // 1 MB - ..Default::default() -}; -let scheduler = IoScheduler::new(config); - -// Calculate optimal buffer size -let buffer_size = calculate_optimal_buffer_size( - 10 * 1024 * 1024, // 10 MB file - 64 * 1024, // base buffer - true, // sequential access - 4, // concurrent requests - StorageMedia::Ssd, - IoLoadLevel::Low, -); -``` - -### Priority Queue - -Priority queue with starvation prevention: - -```rust -use rustfs_io_core::{IoPriorityQueue, IoPriority, IoQueueStatus}; - -let queue = IoPriorityQueue::<()>::new(100); - -// Enqueue request -let request_id = queue.enqueue(IoPriority::High, (), 1024); - -// Dequeue request -if let Some((priority, data)) = queue.dequeue() { - println!("Processing priority {:?} request", priority); -} - -// Check queue status -let status = queue.status(); -println!("High priority waiting: {}", status.high_priority_waiting); -``` - ### Backpressure Control System overload protection: @@ -148,71 +101,23 @@ let stats = optimizer.stats(); println!("Locks acquired: {}", stats.total_acquired()); ``` -### Timeout Wrapper +### Progress Tracking -Dynamic timeout calculation: +Byte progress and staleness for long-running operations: ```rust -use rustfs_io_core::{RequestTimeoutWrapper, TimeoutConfig}; +use rustfs_io_core::OperationProgress; use std::time::Duration; -let config = TimeoutConfig { - base_timeout: Duration::from_secs(5), - timeout_per_mb: Duration::from_millis(100), - max_timeout: Duration::from_secs(300), - ..Default::default() -}; -let wrapper = RequestTimeoutWrapper::new(config); +let progress = OperationProgress::new(Some(1000), Duration::from_secs(5)); -// Calculate operation timeout -let timeout = wrapper.calculate_timeout(10 * 1024 * 1024); // 10 MB -``` - -## Buffer Size Calculation - -Multiple buffer size calculation functions are provided: - -```rust -use rustfs_io_core::{ - get_concurrency_aware_buffer_size, - get_advanced_buffer_size, - get_buffer_size_for_media, - calculate_optimal_buffer_size, - KI_B, MI_B, -}; -use rustfs_io_core::io_profile::StorageMedia; - -// Basic calculation -let size1 = get_concurrency_aware_buffer_size(1024 * 1024, 64 * 1024); - -// Advanced calculation (considering access pattern) -let size2 = get_advanced_buffer_size(10 * 1024 * 1024, 64 * 1024, true); - -// Media type optimization -let size3 = get_buffer_size_for_media(64 * 1024, StorageMedia::Ssd); - -// Comprehensive calculation -let size4 = calculate_optimal_buffer_size( - 100 * 1024 * 1024, // 100 MB file - 64 * 1024, // base buffer - true, // sequential access - 4, // concurrent requests - StorageMedia::Nvme, - IoLoadLevel::Low, -); +progress.update(500); +assert_eq!(progress.progress_percent(), Some(50.0)); +assert!(!progress.is_stale()); ``` ## Configuration -### Environment Variables - -| Variable | Description | Default | -|----------|-------------|---------| -| `RUSTFS_MAX_CONCURRENT_READS` | Max concurrent reads | 64 | -| `RUSTFS_BASE_BUFFER_SIZE` | Base buffer size | 65536 | -| `RUSTFS_MAX_BUFFER_SIZE` | Max buffer size | 1048576 | -| `RUSTFS_IO_TIMEOUT_SECS` | I/O timeout seconds | 30 | - ### Code Configuration ```rust @@ -240,12 +145,11 @@ rustfs-io-core/ ├── src/ │ ├── lib.rs # Module entry │ ├── config.rs # Configuration types -│ ├── scheduler.rs # I/O scheduler -│ ├── io_priority_queue.rs # Priority queue +│ ├── pool.rs # Tiered buffer pool │ ├── backpressure.rs # Backpressure control │ ├── deadlock_detector.rs # Deadlock detection │ ├── lock_optimizer.rs # Lock optimization -│ ├── timeout_wrapper.rs # Timeout wrapper +│ ├── progress.rs # Operation progress tracking │ └── io_profile.rs # I/O profile └── Cargo.toml ``` @@ -254,21 +158,15 @@ rustfs-io-core/ ```bash # Run all tests -cargo test --package rustfs-io-core +cargo nextest run --package rustfs-io-core # Run specific tests -cargo test --package rustfs-io-core --lib scheduler - -# Run benchmarks -cargo bench --package rustfs-io-core +cargo nextest run --package rustfs-io-core -E 'test(backpressure)' ``` ## Documentation - [API Documentation](https://docs.rs/rustfs-io-core) -- [I/O Scheduler Design](./docs/scheduler-design.md) -- [Backpressure Control Design](./docs/backpressure-design.md) -- [Deadlock Detection Algorithm](./docs/deadlock-detection.md) ## Related Modules diff --git a/crates/io-core/README_zh.md b/crates/io-core/README_zh.md index 7471d86c5..9840c1ecd 100644 --- a/crates/io-core/README_zh.md +++ b/crates/io-core/README_zh.md @@ -23,71 +23,20 @@ ## 📖 概述 -**rustfs-io-core** 是 [RustFS](https://rustfs.com) 分布式对象存储系统的核心 I/O 调度模块。它提供了： +**rustfs-io-core** 是 [RustFS](https://rustfs.com) 分布式对象存储系统的共享 I/O 基础组件。它提供了： -- **I/O 调度器**：自适应缓冲区大小计算和负载管理 -- **优先级队列**：支持饥饿预防的请求优先级调度 +- **缓冲池**：分级复用的 `BytesPool` +- **存储画像**：存储介质与访问模式模型（`io_profile`） +- **调度配置**：存储层投影使用的 `IoSchedulerConfig` / `IoPriorityQueueConfig` - **背压控制**：系统过载保护和优雅降级 - **死锁检测**：基于等待图的死锁检测算法 - **锁优化**：自适应自旋锁优化 -- **超时包装器**：动态超时计算和操作进度追踪 +- **进度追踪**：长耗时操作的字节进度与停滞判定 + +调度算法本身位于 `rustfs/src/storage/concurrency/io_schedule.rs`；本 crate 只承载它投影使用的配置形状，不是第二套实现。 ## ✨ 核心功能 -### I/O 调度器 (IoScheduler) - -自适应 I/O 调度，根据文件大小、访问模式和系统负载动态调整缓冲区大小： - -```rust -use rustfs_io_core::{IoScheduler, IoSchedulerConfig, IoLoadLevel}; -use rustfs_io_core::io_profile::{StorageMedia, AccessPattern}; - -// 创建调度器 -let config = IoSchedulerConfig { - max_concurrent_reads: 64, - base_buffer_size: 64 * 1024, // 64 KB - max_buffer_size: 1024 * 1024, // 1 MB - ..Default::default() -}; -let scheduler = IoScheduler::new(config); - -// 计算最优缓冲区大小 -let buffer_size = scheduler.calculate_buffer_size( - 10 * 1024 * 1024, // 10 MB 文件 - true, // 顺序访问 - StorageMedia::Ssd, - IoLoadLevel::Low, -); -println!("缓冲区大小: {} bytes", buffer_size); -``` - -### 优先级队列 (IoPriorityQueue) - -支持饥饿预防的优先级队列： - -```rust -use rustfs_io_core::{IoPriorityQueue, IoPriority, IoQueueStatus}; - -let queue = IoPriorityQueue::<()>::new(100); - -// 入队请求 -let request_id = queue.enqueue( - IoPriority::High, - (), // 请求数据 - 1024, // 请求大小 -); - -// 出队请求 -if let Some((priority, data)) = queue.dequeue() { - println!("处理优先级 {:?} 的请求", priority); -} - -// 检查队列状态 -let status = queue.status(); -println!("高优先级等待: {}", status.high_priority_waiting); -println!("低优先级等待: {}", status.low_priority_waiting); -``` - ### 背压控制 (BackpressureMonitor) 系统过载保护： @@ -165,78 +114,23 @@ let stats = optimizer.stats(); println!("获取锁次数: {}", stats.locks_acquired.load(std::sync::atomic::Ordering::Relaxed)); ``` -### 超时包装器 (RequestTimeoutWrapper) +### 进度追踪 (OperationProgress) -动态超时计算： +长耗时操作的字节进度与停滞判定： ```rust -use rustfs_io_core::{RequestTimeoutWrapper, TimeoutConfig}; +use rustfs_io_core::OperationProgress; use std::time::Duration; -let config = TimeoutConfig { - base_timeout: Duration::from_secs(5), - timeout_per_mb: Duration::from_millis(100), - max_timeout: Duration::from_secs(300), - ..Default::default() -}; -let wrapper = RequestTimeoutWrapper::new(config); +let progress = OperationProgress::new(Some(1000), Duration::from_secs(5)); -// 计算操作超时 -let timeout = wrapper.calculate_timeout(10 * 1024 * 1024); // 10 MB -println!("超时时间: {:?}", timeout); - -// 执行带超时的操作 -let result = wrapper.execute_with_timeout(async { - // 异步操作 - Ok::<_, std::io::Error>(()) -}, timeout).await; -``` - -## 📊 缓冲区大小计算 - -模块提供了多种缓冲区大小计算函数： - -```rust -use rustfs_io_core::{ - get_concurrency_aware_buffer_size, - get_advanced_buffer_size, - get_buffer_size_for_media, - calculate_optimal_buffer_size, - KI_B, MI_B, -}; -use rustfs_io_core::io_profile::StorageMedia; - -// 基础计算 -let size1 = get_concurrency_aware_buffer_size(1024 * 1024, 64 * 1024); - -// 高级计算（考虑访问模式） -let size2 = get_advanced_buffer_size(10 * 1024 * 1024, 64 * 1024, true); - -// 媒体类型优化 -let size3 = get_buffer_size_for_media(64 * 1024, StorageMedia::Ssd); - -// 综合计算 -let size4 = calculate_optimal_buffer_size( - 100 * 1024 * 1024, // 100 MB 文件 - 64 * 1024, // 基础缓冲区 - true, // 顺序访问 - 4, // 并发请求数 - StorageMedia::Nvme, - IoLoadLevel::Low, -); +progress.update(500); +assert_eq!(progress.progress_percent(), Some(50.0)); +assert!(!progress.is_stale()); ``` ## 🔧 配置 -### 环境变量 - -| 变量名 | 描述 | 默认值 | -|--------|------|--------| -| `RUSTFS_MAX_CONCURRENT_READS` | 最大并发读数 | 64 | -| `RUSTFS_BASE_BUFFER_SIZE` | 基础缓冲区大小 | 65536 | -| `RUSTFS_MAX_BUFFER_SIZE` | 最大缓冲区大小 | 1048576 | -| `RUSTFS_IO_TIMEOUT_SECS` | I/O 超时秒数 | 30 | - ### 代码配置 ```rust @@ -264,12 +158,11 @@ rustfs-io-core/ ├── src/ │ ├── lib.rs # 模块入口 │ ├── config.rs # 配置类型 -│ ├── scheduler.rs # I/O 调度器 -│ ├── io_priority_queue.rs # 优先级队列 +│ ├── pool.rs # 分级缓冲池 │ ├── backpressure.rs # 背压控制 │ ├── deadlock_detector.rs # 死锁检测 │ ├── lock_optimizer.rs # 锁优化 -│ ├── timeout_wrapper.rs # 超时包装器 +│ ├── progress.rs # 操作进度追踪 │ └── io_profile.rs # I/O 配置文件 └── Cargo.toml ``` @@ -278,21 +171,15 @@ rustfs-io-core/ ```bash # 运行所有测试 -cargo test --package rustfs-io-core +cargo nextest run --package rustfs-io-core # 运行特定测试 -cargo test --package rustfs-io-core --lib scheduler - -# 运行基准测试 -cargo bench --package rustfs-io-core +cargo nextest run --package rustfs-io-core -E 'test(backpressure)' ``` ## 📚 文档 - [API 文档](https://docs.rs/rustfs-io-core) -- [I/O 调度器设计](./docs/scheduler-design.md) -- [背压控制原理](./docs/backpressure-design.md) -- [死锁检测算法](./docs/deadlock-detection.md) ## 🔗 相关模块 diff --git a/crates/io-core/examples/scheduler_example.rs b/crates/io-core/examples/scheduler_example.rs deleted file mode 100644 index 00fa560ae..000000000 --- a/crates/io-core/examples/scheduler_example.rs +++ /dev/null @@ -1,190 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Example demonstrating I/O scheduler usage. - -use rustfs_io_core::io_profile::StorageMedia; -use rustfs_io_core::{ - BackpressureMonitor, BackpressureState, DeadlockDetector, IoLoadLevel, IoScheduler, IoSchedulerConfig, KI_B, LockOptimizer, - LockType, MI_B, calculate_optimal_buffer_size, get_buffer_size_for_media, -}; -use std::time::Duration; - -fn main() { - println!("=== rustfs-io-core Example ===\n"); - - // 1. I/O scheduler example - io_scheduler_example(); - - // 2. Buffer size calculation example - buffer_size_example(); - - // 3. Backpressure control example - backpressure_example(); - - // 4. Deadlock detection example - deadlock_detection_example(); - - // 5. Lock optimizer example - lock_optimizer_example(); -} - -fn io_scheduler_example() { - println!("--- I/O Scheduler ---"); - - // Create scheduler with configuration - let config = IoSchedulerConfig { - max_concurrent_reads: 64, - base_buffer_size: 64 * KI_B, - max_buffer_size: MI_B, - ..Default::default() - }; - let scheduler = IoScheduler::new(config); - - println!(" Max concurrent reads: {}", scheduler.config().max_concurrent_reads); - println!(" Base buffer size: {} KB", scheduler.config().base_buffer_size / KI_B); - println!(" Max buffer size: {} KB", scheduler.config().max_buffer_size / KI_B); - - // Calculate buffer sizes for different scenarios - let scenarios = [ - ("Small file", 10 * KI_B as i64, true, StorageMedia::Ssd), - ("Medium file", MI_B as i64, true, StorageMedia::Ssd), - ("Large sequential", 100 * MI_B as i64, true, StorageMedia::Ssd), - ("Large random", 100 * MI_B as i64, false, StorageMedia::Ssd), - ("NVMe large", 100 * MI_B as i64, true, StorageMedia::Nvme), - ("HDD large", 100 * MI_B as i64, true, StorageMedia::Hdd), - ]; - - for (name, size, sequential, media) in scenarios { - let buffer = calculate_optimal_buffer_size(size, 64 * KI_B, sequential, 4, media, IoLoadLevel::Low); - println!(" {}: {} bytes ({} KB)", name, buffer, buffer / KI_B); - } - - println!(); -} - -fn buffer_size_example() { - println!("--- Buffer Size Calculation ---"); - - // Comprehensive calculation - let size1 = calculate_optimal_buffer_size(10 * MI_B as i64, 64 * KI_B, true, 4, StorageMedia::Ssd, IoLoadLevel::Low); - println!(" Comprehensive (10MB, sequential, SSD): {} KB", size1 / KI_B); - - // Media type optimization - let media_types = [ - StorageMedia::Nvme, - StorageMedia::Ssd, - StorageMedia::Hdd, - StorageMedia::Unknown, - ]; - for media in media_types { - let size = get_buffer_size_for_media(64 * KI_B, media); - println!(" {} optimized: {} KB", media.as_str(), size / KI_B); - } - - println!(); -} - -fn backpressure_example() { - println!("--- Backpressure Control ---"); - - let monitor = BackpressureMonitor::with_defaults(); - - // Check initial state - let state = monitor.state(); - let state_str = match state { - BackpressureState::Normal => "Normal", - BackpressureState::Warning => "Warning", - BackpressureState::Critical => "Critical", - }; - println!(" Initial state: {}", state_str); - - // Check if active - let is_active = monitor.is_active(); - println!(" Backpressure active: {}", is_active); - - // Try to acquire permit - if monitor.try_acquire() { - println!(" Successfully acquired permit"); - monitor.release(); - println!(" Released permit"); - } - - // View statistics - println!(" Total processed: {}", monitor.total_processed()); - println!(" Total rejected: {}", monitor.total_rejected()); - - println!(); -} - -fn deadlock_detection_example() { - println!("--- Deadlock Detection ---"); - - let detector = DeadlockDetector::with_defaults(); - - // Register locks - let mutex1 = detector.register_lock(LockType::Mutex); - let mutex2 = detector.register_lock(LockType::Mutex); - println!(" Registered locks: mutex1={}, mutex2={}", mutex1, mutex2); - - // Simulate normal operation - detector.record_acquire(mutex1, 1); // Thread 1 acquires mutex1 - detector.record_acquire(mutex2, 2); // Thread 2 acquires mutex2 - println!(" Normal operation: no deadlock"); - - // Detect deadlock - if detector.detect_deadlock().is_none() { - println!(" Detection result: no deadlock"); - } - - // Simulate deadlock scenario - detector.record_wait(mutex2, 1); // Thread 1 waits for mutex2 - detector.record_wait(mutex1, 2); // Thread 2 waits for mutex1 - - // Detect deadlock - if let Some(deadlock) = detector.detect_deadlock() { - println!(" Detection result: deadlock found {:?}", deadlock); - } - - // Cleanup - detector.unregister_lock(mutex1); - detector.unregister_lock(mutex2); - println!(); -} - -fn lock_optimizer_example() { - println!("--- Lock Optimizer ---"); - - let optimizer = LockOptimizer::with_defaults(); - - // Simulate lock operations - for _i in 0..5 { - optimizer.on_acquire(); - // Simulate work - std::thread::sleep(Duration::from_millis(10)); - optimizer.on_release(Duration::from_millis(10)); - } - - // View statistics - let stats = optimizer.stats(); - let acquired = stats.total_acquired(); - let avg_hold = stats.avg_hold_time(); - let contention = stats.contention_rate(); - - println!(" Locks acquired: {}", acquired); - println!(" Average hold time: {:?}", avg_hold); - println!(" Contention rate: {:.2}%", contention * 100.0); - - println!(); -} diff --git a/crates/io-core/src/bufreader_optimizer.rs b/crates/io-core/src/bufreader_optimizer.rs deleted file mode 100644 index 26ec968d5..000000000 --- a/crates/io-core/src/bufreader_optimizer.rs +++ /dev/null @@ -1,227 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! BufReader layer optimizer for minimizing redundant buffering layers. -//! -//! This module provides optimization for BufReader usage in data paths, -//! including layer count limiting and dynamic buffer size adjustment. - -use std::sync::atomic::{AtomicU64, Ordering}; - -/// BufReader optimization configuration. -#[derive(Debug, Clone)] -pub struct BufReaderConfig { - /// Maximum number of nested BufReader layers (default: 2) - pub max_layers: u32, - - /// Buffer size for small files (default: 8KB) - pub small_file_buffer: usize, - - /// Buffer size for large files (default: 64KB) - pub large_file_buffer: usize, - - /// Threshold for large file classification (default: 1MB) - pub large_file_threshold: usize, -} - -impl Default for BufReaderConfig { - fn default() -> Self { - Self { - max_layers: 2, - small_file_buffer: 8 * 1024, // 8KB - large_file_buffer: 64 * 1024, // 64KB - large_file_threshold: 1024 * 1024, // 1MB - } - } -} - -/// BufReader optimization statistics. -#[derive(Debug, Default)] -pub struct BufReaderStats { - /// Total number of readers created - pub total_readers: AtomicU64, - - /// Number of redundant layers eliminated - pub eliminated_layers: AtomicU64, - - /// Number of buffer size adjustments - pub buffer_size_adjustments: AtomicU64, -} - -/// BufReader layer optimizer. -/// -/// Analyzes and optimizes BufReader nesting in data paths, -/// dynamically adjusting buffer sizes based on data characteristics. -pub struct BufReaderOptimizer { - config: BufReaderConfig, - stats: BufReaderStats, -} - -impl BufReaderOptimizer { - /// Create a new BufReader optimizer with the given configuration. - pub fn new(config: BufReaderConfig) -> Self { - Self { - config, - stats: BufReaderStats::default(), - } - } - - /// Create a new BufReader optimizer with default configuration. - pub fn with_defaults() -> Self { - Self::new(BufReaderConfig::default()) - } - - /// Calculate the optimal buffer size based on data size. - /// - /// Returns the appropriate buffer size based on whether the data - /// is classified as a small or large file. - pub fn optimal_buffer_size(&self, data_size: Option) -> usize { - match data_size { - Some(size) if size >= self.config.large_file_threshold => self.config.large_file_buffer, - Some(_) => self.config.small_file_buffer, - None => self.config.small_file_buffer, - } - } - - /// Optimize a reader by wrapping it with an appropriately sized BufReader. - /// - /// This method applies the optimal buffer size based on the expected - /// data size and tracks statistics. - pub fn optimize(&self, reader: R, data_size: Option) -> tokio::io::BufReader { - let buffer_size = self.optimal_buffer_size(data_size); - self.stats.total_readers.fetch_add(1, Ordering::Relaxed); - tokio::io::BufReader::with_capacity(buffer_size, reader) - } - - /// Get the statistics for this optimizer. - pub fn stats(&self) -> &BufReaderStats { - &self.stats - } - - /// Get the configuration for this optimizer. - pub fn config(&self) -> &BufReaderConfig { - &self.config - } -} - -/// Marker trait for buffered sources. -/// -/// Types implementing this trait are considered already buffered -/// and should not be wrapped with additional BufReader layers. -pub trait BufferedSource: tokio::io::AsyncRead {} - -impl BufReaderOptimizer { - /// Check if a reader is already a buffered source. - /// - /// Returns true if the reader implements `BufferedSource`, - /// indicating it should not be wrapped with BufReader. - pub fn is_buffered_source(&self, _reader: &R) -> bool { - true - } - - /// Eliminate redundant BufReader layers if possible. - /// - /// This method attempts to reduce the nesting depth of BufReader - /// layers to improve performance. - pub fn eliminate_redundant_layers(&self, reader: R) -> R { - // For now, just return the reader as-is - // Future implementation could detect and unwrap nested BufReaders - self.stats.eliminated_layers.fetch_add(0, Ordering::Relaxed); - reader - } -} - -#[cfg(test)] -mod tests { - use super::*; - use tokio::io::AsyncReadExt; - - #[test] - fn test_default_config() { - let config = BufReaderConfig::default(); - assert_eq!(config.max_layers, 2); - assert_eq!(config.small_file_buffer, 8 * 1024); - assert_eq!(config.large_file_buffer, 64 * 1024); - assert_eq!(config.large_file_threshold, 1024 * 1024); - } - - #[test] - fn test_optimal_buffer_size_small_file() { - let optimizer = BufReaderOptimizer::with_defaults(); - - // Small file (< 1MB) - assert_eq!(optimizer.optimal_buffer_size(Some(100)), 8 * 1024); - assert_eq!(optimizer.optimal_buffer_size(Some(1024)), 8 * 1024); - assert_eq!(optimizer.optimal_buffer_size(Some(512 * 1024)), 8 * 1024); - } - - #[test] - fn test_optimal_buffer_size_large_file() { - let optimizer = BufReaderOptimizer::with_defaults(); - - // Large file (>= 1MB) - assert_eq!(optimizer.optimal_buffer_size(Some(1024 * 1024)), 64 * 1024); - assert_eq!(optimizer.optimal_buffer_size(Some(10 * 1024 * 1024)), 64 * 1024); - } - - #[test] - fn test_optimal_buffer_size_unknown() { - let optimizer = BufReaderOptimizer::with_defaults(); - - // Unknown size - assert_eq!(optimizer.optimal_buffer_size(None), 8 * 1024); - } - - #[tokio::test] - async fn test_optimize_creates_bufreader() { - let optimizer = BufReaderOptimizer::with_defaults(); - let data = vec![1u8, 2, 3, 4, 5]; - let cursor = std::io::Cursor::new(data.clone()); - - let mut reader = optimizer.optimize(cursor, Some(5)); - - let mut buf = vec![0u8; 5]; - let n = reader.read(&mut buf).await.unwrap(); - - assert_eq!(n, 5); - assert_eq!(buf, data); - } - - #[test] - fn test_stats_tracking() { - let optimizer = BufReaderOptimizer::with_defaults(); - - assert_eq!(optimizer.stats().total_readers.load(Ordering::Relaxed), 0); - - let cursor = std::io::Cursor::new(vec![1u8, 2, 3]); - let _reader = optimizer.optimize(cursor, Some(3)); - - assert_eq!(optimizer.stats().total_readers.load(Ordering::Relaxed), 1); - } - - #[test] - fn test_custom_config() { - let config = BufReaderConfig { - max_layers: 3, - small_file_buffer: 4 * 1024, - large_file_buffer: 128 * 1024, - large_file_threshold: 2 * 1024 * 1024, - }; - - let optimizer = BufReaderOptimizer::new(config); - - assert_eq!(optimizer.optimal_buffer_size(Some(1024 * 1024)), 4 * 1024); - assert_eq!(optimizer.optimal_buffer_size(Some(3 * 1024 * 1024)), 128 * 1024); - } -} diff --git a/crates/io-core/src/direct_io.rs b/crates/io-core/src/direct_io.rs deleted file mode 100644 index 62e9834c8..000000000 --- a/crates/io-core/src/direct_io.rs +++ /dev/null @@ -1,332 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Aligned pread-based file reader. -//! -//! This module provides an aligned, position-based file reader that uses -//! `pread`/`FileExt::read_at` for I/O operations. It performs reads at -//! 512-byte-aligned offsets and sizes, making it suitable as a foundation -//! for workloads where alignment matters. -//! -//! Note: This reader does **not** set the `O_DIRECT` flag and therefore does -//! not bypass the OS page cache. It is an aligned `pread`-based reader, not -//! true Direct I/O. To implement true O_DIRECT on Linux, the file must be -//! opened with `O_DIRECT` via `libc::open`. -//! -//! # Platform Support -//! -//! The `read_at` implementation is only available on Unix-like platforms. -//! On other platforms, this reader will return an error. - -use std::io::{self}; -use std::pin::Pin; -use std::task::{Context, Poll}; -use tokio::io::{AsyncRead, ReadBuf}; - -/// Errors that can occur during aligned pread operations. -#[derive(Debug, Clone)] -pub enum AlignedPreadError { - /// Platform doesn't support `read_at`-based I/O - UnsupportedPlatform, - /// File descriptor doesn't support this reader - UnsupportedFile, - /// I/O error occurred - Io(String), - /// Invalid alignment (reads require 512-byte-aligned offset and size) - AlignmentError { offset: u64, size: usize }, -} - -impl std::fmt::Display for AlignedPreadError { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - match self { - Self::UnsupportedPlatform => write!(f, "Aligned pread not supported on this platform"), - Self::UnsupportedFile => write!(f, "File doesn't support this reader"), - Self::Io(msg) => write!(f, "I/O error: {}", msg), - Self::AlignmentError { offset, size } => { - write!(f, "Alignment error: offset={}, size={}", offset, size) - } - } - } -} - -impl std::error::Error for AlignedPreadError {} - -impl From for AlignedPreadError { - fn from(err: io::Error) -> Self { - Self::Io(err.to_string()) - } -} - -/// Aligned pread-based file reader for Unix platforms. -/// -/// This reader performs I/O using `pread`/`FileExt::read_at` at -/// 512-byte-aligned offsets and sizes, without modifying the file's -/// current position. -/// -/// **Note:** This reader does **not** set the `O_DIRECT` flag and therefore -/// does **not** bypass the OS page cache. It is an aligned `pread`-based -/// reader. To implement true O_DIRECT, the file must be opened with -/// `O_DIRECT` via `libc::open`. -/// -/// # Platform Support -/// -/// Only available on Linux (uses `FileExt::read_at`). On other platforms, -/// use `BytesBufferedReader` instead. -/// -/// # Alignment Requirements -/// -/// Reads have strict alignment requirements: -/// - File offset must be aligned to 512 bytes -/// - Buffer size must be a multiple of 512 bytes -/// - Buffer address must be aligned (handled internally) -/// -/// # Example -/// -/// ```ignore -/// use rustfs_io_core::AlignedPreadReader; -/// -/// // Linux only -/// #[cfg(target_os = "linux")] -/// let reader = AlignedPreadReader::new(file, offset, size)?; -/// ``` -#[cfg(target_os = "linux")] -pub struct AlignedPreadReader { - /// Underlying file handle used for aligned pread I/O - file: std::fs::File, - /// Current read position - pos: u64, - /// Remaining bytes to read - remaining: usize, - /// Buffer for aligned reads - buffer: Vec, - /// Current position in the buffer - buffer_pos: usize, - /// Amount of data in the buffer - buffer_len: usize, -} - -#[cfg(target_os = "linux")] -impl AlignedPreadReader { - /// Alignment requirement for reads (512 bytes for most systems) - pub const ALIGNMENT: usize = 512; - - /// Create a new aligned pread-based reader. - /// - /// # Arguments - /// - /// * `file` - File to read from - /// * `offset` - Starting offset in the file (must be 512-byte aligned) - /// * `size` - Number of bytes to read (must be 512-byte aligned) - /// - /// # Returns - /// - /// An `AlignedPreadReader` that reads the file at the given offset. - /// - /// # Errors - /// - /// Returns an error if offset or size are not 512-byte aligned. - pub fn new(file: std::fs::File, offset: u64, size: usize) -> Result { - // Check alignment - if !offset.is_multiple_of(Self::ALIGNMENT as u64) { - return Err(AlignedPreadError::AlignmentError { offset, size }); - } - if !size.is_multiple_of(Self::ALIGNMENT) { - return Err(AlignedPreadError::AlignmentError { offset, size }); - } - - Ok(Self { - file, - pos: offset, - remaining: size, - buffer: Vec::new(), - buffer_pos: 0, - buffer_len: 0, - }) - } - - /// Read a chunk of data using aligned pread. - /// - /// This method performs aligned reads and handles the buffering required - /// by this aligned pread implementation. It does not use `O_DIRECT`. - fn read_chunk(&mut self, buf: &mut [u8]) -> io::Result { - // If buffer is exhausted, read more data - if self.buffer_pos >= self.buffer_len { - if self.remaining == 0 { - return Ok(0); - } - - // Allocate aligned buffer - let chunk_size = (self.remaining).min(64 * 1024); // 64KB chunks - let aligned_size = chunk_size.div_ceil(Self::ALIGNMENT) * Self::ALIGNMENT; - - self.buffer = vec![0u8; aligned_size]; - - // Use pread for atomic read at position (no file offset modification) - use std::os::unix::fs::FileExt; - let n = self.file.read_at(&mut self.buffer, self.pos)?; - - self.buffer_pos = 0; - self.buffer_len = n; - self.pos += n as u64; - self.remaining -= n; - - if n == 0 { - return Ok(0); - } - } - - // Copy from buffer to user buffer - let available = self.buffer_len - self.buffer_pos; - let to_copy = buf.len().min(available); - buf[..to_copy].copy_from_slice(&self.buffer[self.buffer_pos..self.buffer_pos + to_copy]); - self.buffer_pos += to_copy; - - Ok(to_copy) - } -} - -#[cfg(target_os = "linux")] -impl AsyncRead for AlignedPreadReader { - fn poll_read(mut self: Pin<&mut Self>, _cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll> { - let filled = buf.filled().len(); - let mut remaining = buf.initialize_unfilled(); - - while !remaining.is_empty() { - match self.read_chunk(remaining) { - Ok(0) => break, - Ok(n) => { - remaining = &mut remaining[n..]; - } - Err(e) => return Poll::Ready(Err(e)), - } - } - - let _n_read = buf.filled().len() - filled; - Poll::Ready(Ok(())) - } -} - -/// Aligned pread reader stub for non-Linux platforms. -/// -/// On non-Linux platforms, `read_at`-based I/O is not available through this -/// type. This stub exists to provide a consistent API across platforms. -#[cfg(not(target_os = "linux"))] -pub struct AlignedPreadReader { - _priv: (), -} - -#[cfg(not(target_os = "linux"))] -impl AlignedPreadReader { - /// Create a new aligned pread reader (not supported on this platform). - /// - /// Always returns an error on non-Linux platforms. - pub fn new(_file: std::fs::File, _offset: u64, _size: usize) -> Result { - Err(AlignedPreadError::UnsupportedPlatform) - } -} - -#[cfg(not(target_os = "linux"))] -impl AsyncRead for AlignedPreadReader { - fn poll_read(self: Pin<&mut Self>, _cx: &mut Context<'_>, _buf: &mut ReadBuf<'_>) -> Poll> { - Poll::Ready(Err(io::Error::new( - io::ErrorKind::Unsupported, - "Aligned pread-based I/O not supported on this platform", - ))) - } -} - -impl std::fmt::Debug for AlignedPreadReader { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - #[cfg(target_os = "linux")] - { - f.debug_struct("AlignedPreadReader") - .field("pos", &self.pos) - .field("remaining", &self.remaining) - .field("buffer_len", &self.buffer_len) - .finish() - } - #[cfg(not(target_os = "linux"))] - { - f.debug_struct("AlignedPreadReader") - .field("platform", &"unsupported") - .finish() - } - } -} - -/// Historical name for aligned pread errors. -#[deprecated(since = "1.0.0-beta.8", note = "use AlignedPreadError; this reader does not set O_DIRECT")] -pub type DirectIoError = AlignedPreadError; - -/// Historical name for the aligned pread-based reader. -#[deprecated(since = "1.0.0-beta.8", note = "use AlignedPreadReader; this reader does not set O_DIRECT")] -pub type DirectIoReader = AlignedPreadReader; - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn test_alignment_check() { - #[cfg(target_os = "linux")] - { - // Valid alignment - let file = std::fs::File::open("/dev/zero").unwrap(); - assert!( - AlignedPreadReader::new(file, 0, 512).is_ok(), - "Should succeed with aligned offset and size" - ); - - let file = std::fs::File::open("/dev/zero").expect("open /dev/zero for alias"); - assert!( - AlignedPreadReader::new(file, 0, 512).is_ok(), - "Should succeed through aligned pread alias" - ); - - // Invalid offset - let file = std::fs::File::open("/dev/zero").unwrap(); - assert!(AlignedPreadReader::new(file, 1, 512).is_err(), "Should fail with unaligned offset"); - - // Invalid size - let file = std::fs::File::open("/dev/zero").unwrap(); - assert!(AlignedPreadReader::new(file, 0, 511).is_err(), "Should fail with unaligned size"); - } - - #[cfg(not(target_os = "linux"))] - { - // Non-Linux should return UnsupportedPlatform - let file = std::fs::File::open(std::env::current_exe().unwrap()).unwrap(); - assert!(matches!( - AlignedPreadReader::new(file, 0, 512), - Err(AlignedPreadError::UnsupportedPlatform) - )); - } - } - - #[test] - #[allow(deprecated)] - fn test_legacy_direct_io_alias() { - #[cfg(target_os = "linux")] - { - let file = std::fs::File::open("/dev/zero").unwrap(); - assert!(DirectIoReader::new(file, 0, 512).is_ok()); - } - - #[cfg(not(target_os = "linux"))] - { - let file = std::fs::File::open(std::env::current_exe().unwrap()).unwrap(); - assert!(matches!(DirectIoReader::new(file, 0, 512), Err(AlignedPreadError::UnsupportedPlatform))); - } - } -} diff --git a/crates/io-core/src/io_priority_queue.rs b/crates/io-core/src/io_priority_queue.rs deleted file mode 100644 index bb7b4cf45..000000000 --- a/crates/io-core/src/io_priority_queue.rs +++ /dev/null @@ -1,381 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! I/O priority queue for scheduling I/O operations. -//! -//! This module provides a priority queue implementation for I/O operations -//! with support for starvation prevention and fair scheduling. - -use crate::config::IoPriorityQueueConfig; -use crate::scheduler::IoPriority; -use std::collections::VecDeque; -use std::time::{Duration, Instant}; - -/// A queued I/O request. -#[derive(Debug, Clone)] -pub struct IoRequest { - /// Request ID. - pub id: u64, - /// Request priority. - pub priority: IoPriority, - /// Request size in bytes. - pub size: usize, - /// Queue time. - pub queued_at: Instant, - /// Whether this is a sequential read. - pub is_sequential: bool, -} - -impl IoRequest { - /// Create a new I/O request. - pub fn new(id: u64, priority: IoPriority, size: usize, is_sequential: bool) -> Self { - Self { - id, - priority, - size, - queued_at: Instant::now(), - is_sequential, - } - } - - /// Get the wait time in the queue. - pub fn wait_time(&self) -> Duration { - self.queued_at.elapsed() - } -} - -/// Queue status for a priority level. -#[derive(Debug, Clone, Default)] -pub struct IoQueueStatus { - /// Number of requests in the queue. - pub count: usize, - /// Total size of all requests. - pub total_size: usize, - /// Oldest request wait time. - pub oldest_wait: Option, - /// Number of requests processed. - pub processed: u64, -} - -impl IoQueueStatus { - /// Create new queue status. - pub fn new() -> Self { - Self::default() - } -} - -/// I/O priority queue. -pub struct IoPriorityQueue { - /// Queue configuration. - config: IoPriorityQueueConfig, - /// High priority queue. - high: VecDeque, - /// Normal priority queue. - normal: VecDeque, - /// Low priority queue. - low: VecDeque, - /// Next request ID. - next_id: u64, - /// Last dequeue time for each priority (for starvation prevention). - last_dequeue: [Option; 3], - /// Statistics for each queue. - stats: [IoQueueStatus; 3], -} - -impl IoPriorityQueue { - /// Create a new priority queue with the given configuration. - pub fn new(config: IoPriorityQueueConfig) -> Self { - Self { - config, - high: VecDeque::with_capacity(100), - normal: VecDeque::with_capacity(500), - low: VecDeque::with_capacity(200), - next_id: 0, - last_dequeue: [None, None, None], - stats: [IoQueueStatus::new(), IoQueueStatus::new(), IoQueueStatus::new()], - } - } - - /// Create with default configuration. - pub fn with_defaults() -> Self { - Self::new(IoPriorityQueueConfig::default()) - } - - /// Get the configuration. - pub fn config(&self) -> &IoPriorityQueueConfig { - &self.config - } - - /// Enqueue a request. - pub fn enqueue(&mut self, priority: IoPriority, size: usize, is_sequential: bool) -> u64 { - let id = self.next_id; - self.next_id += 1; - - let request = IoRequest::new(id, priority, size, is_sequential); - - match priority { - IoPriority::High => { - if self.high.len() < self.config.high_capacity { - self.high.push_back(request); - } - } - IoPriority::Normal => { - if self.normal.len() < self.config.normal_capacity { - self.normal.push_back(request); - } - } - IoPriority::Low => { - if self.low.len() < self.config.low_capacity { - self.low.push_back(request); - } - } - } - - id - } - - /// Dequeue the next request. - /// - /// Uses weighted fair queuing with starvation prevention. - pub fn dequeue(&mut self) -> Option { - let now = Instant::now(); - - // Check for starvation: if a lower priority queue hasn't been served in a while, - // give it priority - let normal_starved = self.is_starved(IoPriority::Normal, now); - let low_starved = self.is_starved(IoPriority::Low, now); - - // Priority order with starvation consideration - // Check conditions first, then dequeue - let dequeue_high = !self.high.is_empty() && !low_starved && !normal_starved; - let dequeue_normal = !self.normal.is_empty() && !low_starved; - let dequeue_low = !self.low.is_empty(); - let dequeue_high_fallback = !self.high.is_empty(); - let dequeue_normal_fallback = !self.normal.is_empty(); - - if dequeue_high { - let request = self.high.pop_front(); - if request.is_some() { - self.last_dequeue[0] = Some(Instant::now()); - self.stats[0].processed += 1; - } - request - } else if dequeue_normal { - let request = self.normal.pop_front(); - if request.is_some() { - self.last_dequeue[1] = Some(Instant::now()); - self.stats[1].processed += 1; - } - request - } else if dequeue_low { - let request = self.low.pop_front(); - if request.is_some() { - self.last_dequeue[2] = Some(Instant::now()); - self.stats[2].processed += 1; - } - request - } else if dequeue_high_fallback { - let request = self.high.pop_front(); - if request.is_some() { - self.last_dequeue[0] = Some(Instant::now()); - self.stats[0].processed += 1; - } - request - } else if dequeue_normal_fallback { - let request = self.normal.pop_front(); - if request.is_some() { - self.last_dequeue[1] = Some(Instant::now()); - self.stats[1].processed += 1; - } - request - } else { - None - } - } - - /// Check if a priority level is starved. - fn is_starved(&self, priority: IoPriority, now: Instant) -> bool { - let idx = match priority { - IoPriority::High => 0, - IoPriority::Normal => 1, - IoPriority::Low => 2, - }; - - if let Some(last) = self.last_dequeue[idx] { - now.duration_since(last) > self.config.starvation_threshold - } else { - false - } - } - - /// Get the total number of queued requests. - pub fn len(&self) -> usize { - self.high.len() + self.normal.len() + self.low.len() - } - - /// Check if the queue is empty. - pub fn is_empty(&self) -> bool { - self.high.is_empty() && self.normal.is_empty() && self.low.is_empty() - } - - /// Get queue status for a priority level. - pub fn status(&self, priority: IoPriority) -> IoQueueStatus { - let (queue, idx) = match priority { - IoPriority::High => (&self.high, 0), - IoPriority::Normal => (&self.normal, 1), - IoPriority::Low => (&self.low, 2), - }; - - let mut status = self.stats[idx].clone(); - status.count = queue.len(); - status.total_size = queue.iter().map(|r| r.size).sum(); - status.oldest_wait = queue.front().map(|r| r.wait_time()); - status - } - - /// Get the total queue status. - pub fn total_status(&self) -> IoQueueStatus { - let mut total = IoQueueStatus::new(); - total.count = self.len(); - total.total_size = self - .high - .iter() - .chain(self.normal.iter()) - .chain(self.low.iter()) - .map(|r| r.size) - .sum(); - total.processed = self.stats.iter().map(|s| s.processed).sum(); - total.oldest_wait = self - .high - .front() - .map(|r| r.wait_time()) - .or_else(|| self.normal.front().map(|r| r.wait_time())) - .or_else(|| self.low.front().map(|r| r.wait_time())); - total - } - - /// Clear all queues. - pub fn clear(&mut self) { - self.high.clear(); - self.normal.clear(); - self.low.clear(); - } - - /// Peek at the next request without removing it. - pub fn peek(&self) -> Option<&IoRequest> { - if !self.high.is_empty() { - self.high.front() - } else if !self.normal.is_empty() { - self.normal.front() - } else { - self.low.front() - } - } -} - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn test_enqueue_dequeue() { - let mut queue = IoPriorityQueue::with_defaults(); - - let id1 = queue.enqueue(IoPriority::High, 1024, true); - let id2 = queue.enqueue(IoPriority::Normal, 2048, false); - let id3 = queue.enqueue(IoPriority::Low, 4096, true); - - assert_eq!(queue.len(), 3); - - // High priority should be dequeued first - let req1 = queue.dequeue().unwrap(); - assert_eq!(req1.id, id1); - assert_eq!(req1.priority, IoPriority::High); - - let req2 = queue.dequeue().unwrap(); - assert_eq!(req2.id, id2); - assert_eq!(req2.priority, IoPriority::Normal); - - let req3 = queue.dequeue().unwrap(); - assert_eq!(req3.id, id3); - assert_eq!(req3.priority, IoPriority::Low); - - assert!(queue.is_empty()); - } - - #[test] - fn test_queue_status() { - let mut queue = IoPriorityQueue::with_defaults(); - - queue.enqueue(IoPriority::High, 1024, true); - queue.enqueue(IoPriority::High, 2048, true); - queue.enqueue(IoPriority::Normal, 4096, false); - - let high_status = queue.status(IoPriority::High); - assert_eq!(high_status.count, 2); - assert_eq!(high_status.total_size, 3072); - - let normal_status = queue.status(IoPriority::Normal); - assert_eq!(normal_status.count, 1); - assert_eq!(normal_status.total_size, 4096); - - let total = queue.total_status(); - assert_eq!(total.count, 3); - assert_eq!(total.total_size, 7168); - } - - #[test] - fn test_queue_capacity() { - let config = IoPriorityQueueConfig { - high_capacity: 2, - normal_capacity: 2, - low_capacity: 2, - ..Default::default() - }; - let mut queue = IoPriorityQueue::new(config); - - queue.enqueue(IoPriority::High, 1024, true); - queue.enqueue(IoPriority::High, 1024, true); - queue.enqueue(IoPriority::High, 1024, true); // Should be dropped - - assert_eq!(queue.status(IoPriority::High).count, 2); - } - - #[test] - fn test_clear() { - let mut queue = IoPriorityQueue::with_defaults(); - - queue.enqueue(IoPriority::High, 1024, true); - queue.enqueue(IoPriority::Normal, 2048, false); - queue.enqueue(IoPriority::Low, 4096, true); - - assert_eq!(queue.len(), 3); - queue.clear(); - assert!(queue.is_empty()); - } - - #[test] - fn test_peek() { - let mut queue = IoPriorityQueue::with_defaults(); - - queue.enqueue(IoPriority::Normal, 2048, false); - queue.enqueue(IoPriority::High, 1024, true); - - let peeked = queue.peek().unwrap(); - assert_eq!(peeked.priority, IoPriority::High); - - // Peek shouldn't remove the item - assert_eq!(queue.len(), 2); - } -} diff --git a/crates/io-core/src/lib.rs b/crates/io-core/src/lib.rs index 67a374ea7..ca5f20c53 100644 --- a/crates/io-core/src/lib.rs +++ b/crates/io-core/src/lib.rs @@ -12,85 +12,39 @@ // See the License for the specific language governing permissions and // limitations under the License. -//! Buffered I/O reader and writer implementations for RustFS. +//! Shared I/O primitives for RustFS. //! -//! This crate provides buffered readers and writers for I/O operations. -//! Prefer `BytesBufferedReader`, `BytesMutWriter`, and `AlignedPreadReader` -//! for new code. Historical `ZeroCopy*` and `DirectIo*` names remain exported -//! for backward compatibility. +//! This crate holds the buffer pool and the concurrency-control primitives +//! that the storage layer builds on: //! -//! # Features -//! -//! - Memory-mapped file reading (mmap-then-copy) on Unix platforms -//! - Bytes-based buffered wrapping -//! - AsyncRead trait implementations -//! - Tiered BytesPool for buffer management -//! - Aligned pread-based reader (NOT true Direct I/O / O_DIRECT) +//! - Tiered `BytesPool` for buffer management +//! - Storage-media and access-pattern profiling (`io_profile`) +//! - Scheduler and priority-queue configuration shapes +//! - Backpressure admission, deadlock detection, lock optimization +//! - Progress tracking for long-running operations //! //! # Example //! //! ```ignore -//! use rustfs_io_core::{BytesBufferedReader, BytesPool}; -//! use bytes::Bytes; +//! use rustfs_io_core::BytesPool; //! -//! // Create from existing bytes (zero-copy) -//! let data = Bytes::from("hello world"); -//! let reader = BytesBufferedReader::from_bytes(data); -//! -//! // Create from file using buffered reads -//! let reader = BytesBufferedReader::from_file_read(&file, 0, 1024).await?; -//! -//! // Use BytesPool //! let pool = BytesPool::new_tiered(); //! let mut buffer = pool.acquire_buffer(8192).await; //! ``` pub mod backpressure; -pub mod bufreader_optimizer; pub mod config; pub mod deadlock_detector; -pub mod direct_io; -pub mod io_priority_queue; pub mod io_profile; pub mod lock_optimizer; pub mod pool; -pub mod reader; -pub mod scheduler; -pub mod shared_memory; -pub mod timeout_wrapper; -pub mod writer; +pub mod progress; -#[cfg(target_os = "linux")] -pub use direct_io::{AlignedPreadError, AlignedPreadReader}; -#[cfg(target_os = "linux")] -#[allow(deprecated)] -pub use direct_io::{DirectIoError, DirectIoReader}; pub use pool::{BytesPool, BytesPoolConfig, BytesPoolMetrics, PooledBuffer}; -#[allow(deprecated)] -pub use reader::ZeroCopyObjectReader; -pub use reader::{BytesBufferedReader, ZeroCopyReadError}; -#[allow(deprecated)] -pub use writer::ZeroCopyObjectWriter; -pub use writer::{BytesMutWriter, ZeroCopyWriteError}; - -// BufReader optimizer exports -pub use bufreader_optimizer::{BufReaderConfig, BufReaderOptimizer, BufReaderStats, BufferedSource}; - -// Shared memory exports -pub use shared_memory::{ArcData, ArcMetadata, SharedMemoryConfig, SharedMemoryPool, SharedMemoryStats}; // Config exports pub use config::{ConfigError, IoPriorityQueueConfig, IoSchedulerConfig}; -// Scheduler exports -pub use scheduler::{ - BandwidthTier, IoLoadLevel, IoLoadMetrics, IoPriority, IoScheduler, IoSchedulingContext, IoStrategy, KI_B, MI_B, - calculate_optimal_buffer_size, get_advanced_buffer_size, get_buffer_size_for_media, get_concurrency_aware_buffer_size, -}; - -// Priority queue exports -pub use io_priority_queue::{IoPriorityQueue, IoQueueStatus, IoRequest}; - // Backpressure exports pub use backpressure::{BackpressureConfig, BackpressureError, BackpressureMonitor, BackpressureState}; @@ -100,8 +54,5 @@ pub use deadlock_detector::{DeadlockDetector, DeadlockDetectorConfig, LockInfo, // Lock optimizer exports pub use lock_optimizer::{LockGuard, LockOptimizeConfig, LockOptimizer, LockStats}; -// Timeout wrapper exports -pub use timeout_wrapper::{ - OperationProgress, RequestTimeoutWrapper, TimeoutConfig, TimeoutError, TimeoutStats, calculate_adaptive_timeout, - estimate_bytes_per_second, -}; +// Progress tracking exports +pub use progress::OperationProgress; diff --git a/crates/io-core/src/progress.rs b/crates/io-core/src/progress.rs new file mode 100644 index 000000000..44e4460dd --- /dev/null +++ b/crates/io-core/src/progress.rs @@ -0,0 +1,138 @@ +// Copyright 2024 RustFS Team +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +//! Progress tracking for long-running I/O operations. +//! +//! Re-exported as `rustfs_concurrency::OperationProgress` for the storage +//! timeout implementation, which uses `is_stale` to tell a slow transfer +//! apart from a stalled one. + +use std::sync::atomic::{AtomicU64, Ordering}; +use std::time::{Duration, Instant}; + +/// Operation progress tracker. +#[derive(Debug)] +pub struct OperationProgress { + /// Total size (if known). + pub total_size: Option, + /// Bytes processed. + bytes_processed: AtomicU64, + /// Last update time. + last_update: std::sync::Mutex, + /// Stale timeout. + stale_timeout: Duration, + /// Start time for transfer rate calculation. + start_time: Instant, +} + +impl OperationProgress { + /// Create new operation progress. + pub fn new(total_size: Option, stale_timeout: Duration) -> Self { + Self { + total_size, + bytes_processed: AtomicU64::new(0), + last_update: std::sync::Mutex::new(Instant::now()), + stale_timeout, + start_time: Instant::now(), + } + } + + /// Update progress. + pub fn update(&self, bytes: u64) { + self.bytes_processed.store(bytes, Ordering::Relaxed); + if let Ok(mut last) = self.last_update.lock() { + *last = Instant::now(); + } + } + + /// Add to progress. + pub fn add(&self, bytes: u64) { + self.bytes_processed.fetch_add(bytes, Ordering::Relaxed); + if let Ok(mut last) = self.last_update.lock() { + *last = Instant::now(); + } + } + + /// Get current progress. + pub fn current(&self) -> u64 { + self.bytes_processed.load(Ordering::Relaxed) + } + + /// Check if progress is stale. + pub fn is_stale(&self) -> bool { + if let Ok(last) = self.last_update.lock() { + last.elapsed() > self.stale_timeout + } else { + false + } + } + + /// Get progress percentage. + pub fn progress_percent(&self) -> Option { + self.total_size.map(|total| { + if total == 0 { + 100.0 + } else { + let processed = self.bytes_processed.load(Ordering::Relaxed); + (processed as f64 / total as f64 * 100.0).min(100.0) + } + }) + } + + /// Get remaining bytes. + pub fn remaining(&self) -> Option { + self.total_size.map(|total| { + let processed = self.bytes_processed.load(Ordering::Relaxed); + total.saturating_sub(processed) + }) + } + + /// Calculate transfer rate in bytes per second. + /// + /// Returns 0 if no time has elapsed or no data transferred. + pub fn transfer_rate(&self) -> u64 { + let processed = self.bytes_processed.load(Ordering::Relaxed); + if processed == 0 { + return 0; + } + + let elapsed = self.start_time.elapsed().as_secs_f64(); + if elapsed > 0.0 { + (processed as f64 / elapsed) as u64 + } else { + 0 + } + } +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn test_operation_progress() { + let progress = OperationProgress::new(Some(1000), Duration::from_secs(5)); + + assert_eq!(progress.current(), 0); + assert_eq!(progress.progress_percent(), Some(0.0)); + + progress.update(500); + assert_eq!(progress.current(), 500); + assert_eq!(progress.progress_percent(), Some(50.0)); + + progress.add(300); + assert_eq!(progress.current(), 800); + assert_eq!(progress.remaining(), Some(200)); + } +} diff --git a/crates/io-core/src/reader.rs b/crates/io-core/src/reader.rs deleted file mode 100644 index 49433af73..000000000 --- a/crates/io-core/src/reader.rs +++ /dev/null @@ -1,412 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Bytes-backed object reader implementation. - -use bytes::Bytes; -use std::io; -use std::pin::Pin; -use std::task::{Context, Poll}; -use tokio::io::{AsyncRead, ReadBuf}; - -/// Errors that can occur during Bytes-backed read operations. -#[derive(Debug, Clone)] -pub enum ZeroCopyReadError { - /// I/O error occurred. - Io(String), - /// Memory mapping error. - Mmap(String), - /// Invalid offset or size. - InvalidRange, -} - -impl std::fmt::Display for ZeroCopyReadError { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - match self { - Self::Io(msg) => write!(f, "I/O error: {}", msg), - Self::Mmap(msg) => write!(f, "Mmap error: {}", msg), - Self::InvalidRange => write!(f, "Invalid offset or size"), - } - } -} - -impl std::error::Error for ZeroCopyReadError {} - -impl From for ZeroCopyReadError { - fn from(err: io::Error) -> Self { - Self::Io(err.to_string()) - } -} - -/// Bytes-backed object reader. -/// -/// `from_bytes` wraps existing `Bytes` without copying, but file constructors -/// copy file data into owned `Bytes` after mmap or normal reads. -/// -/// # Example -/// -/// ```ignore -/// use bytes::Bytes; -/// use rustfs_io_core::BytesBufferedReader; -/// -/// // Create from bytes without copying the `Bytes` buffer -/// let data = Bytes::from("hello world"); -/// let reader = BytesBufferedReader::from_bytes(data); -/// -/// // Read using AsyncRead trait -/// let mut buf = vec![0u8; 1024]; -/// let n = reader.read(&mut buf[..]).await?; -/// ``` -pub struct BytesBufferedReader { - /// Internal data source (could be mmap or owned bytes) - data: Bytes, - /// Current read position - pos: usize, -} - -/// Historical name for the bytes-backed object reader. -#[deprecated( - since = "1.0.0-beta.8", - note = "use BytesBufferedReader; file constructors copy into owned Bytes" -)] -pub type ZeroCopyObjectReader = BytesBufferedReader; - -impl BytesBufferedReader { - /// Create a reader from existing bytes. - /// - /// This is a true zero-copy operation - the Bytes are wrapped - /// without any allocation or copying. - /// - /// # Arguments - /// - /// * `data` - Bytes to wrap - /// - /// # Example - /// - /// ```ignore - /// let data = Bytes::from("hello world"); - /// let reader = BytesBufferedReader::from_bytes(data); - /// ``` - pub fn from_bytes(data: Bytes) -> Self { - Self { data, pos: 0 } - } - - /// Create a Bytes-backed reader from a file using mmap-then-copy. - /// - /// This maps the requested file range and copies it into owned `Bytes` - /// before returning. It does not expose the mmap as a zero-copy buffer. - /// - /// # Arguments - /// - /// * `path` - Path to the file to memory map - /// * `offset` - Offset within the file to start reading - /// * `size` - Number of bytes to read - /// - /// # Returns - /// - /// A reader backed by copied file data. - /// - /// # Errors - /// - /// Returns an error if the file cannot be memory mapped. - /// - /// # Example - /// - /// ```ignore - /// let reader = BytesBufferedReader::from_file_mmap_path("large_file.bin", 0, 1024).await?; - /// ``` - #[cfg(unix)] - // SAFETY: The mmap is created from a read-only file handle for the - // caller-provided range, then copied into owned `Bytes` before the file and - // mapping are dropped. - #[allow(unsafe_code)] - pub async fn from_file_mmap_path(path: &std::path::Path, offset: u64, size: usize) -> Result { - use memmap2::MmapOptions; - - let path = path.to_path_buf(); - let (offset, size) = (offset, size); - - tokio::task::spawn_blocking(move || { - // Open the file in sync context - let std_file = std::fs::File::open(&path).map_err(|e| ZeroCopyReadError::Io(e.to_string()))?; - - // SAFETY: `std_file` remains open while the mapping is created and - // copied, and the mapped bytes are not exposed beyond this closure. - let mmap = unsafe { MmapOptions::new().offset(offset).len(size).map(&std_file) } - .map_err(|e| ZeroCopyReadError::Mmap(e.to_string()))?; - - // Convert to Bytes (this is a copy, but only done once) - Ok(Self { - data: Bytes::copy_from_slice(&mmap), - pos: 0, - }) - }) - .await - .map_err(|e| ZeroCopyReadError::Io(e.to_string()))? - } - - /// Create a Bytes-backed reader from a file using normal reads. - /// - /// This path reads the requested range into an owned buffer and wraps it in - /// `Bytes`. It does not perform mmap or zero-copy file I/O. - /// - /// # Arguments - /// - /// * `file` - File to read from - /// * `offset` - Offset within the file to start reading - /// * `size` - Number of bytes to map - /// - /// # Returns - /// - /// A reader backed by copied file data. - /// - /// # Errors - /// - /// Returns an error if the file cannot be read. - /// - /// # Example - /// - /// ```ignore - /// let file = tokio::fs::File::open("large_file.bin").await?; - /// let reader = BytesBufferedReader::from_file_read(&file, 0, 1024).await?; - /// ``` - #[cfg(unix)] - pub async fn from_file_read(file: &tokio::fs::File, offset: u64, size: usize) -> Result { - use tokio::io::{AsyncReadExt, AsyncSeekExt, SeekFrom}; - - let mut cloned = file.try_clone().await?; - cloned.seek(SeekFrom::Start(offset)).await?; - - let mut buffer = vec![0u8; size]; - cloned.read_exact(&mut buffer).await?; - - Ok(Self { - data: Bytes::from(buffer), - pos: 0, - }) - } - - /// Create a Bytes-backed reader from a file (non-Unix fallback). - /// - /// On platforms that don't support mmap, this falls back to regular file I/O. - #[cfg(not(unix))] - pub async fn from_file_read(file: &tokio::fs::File, offset: u64, size: usize) -> Result { - use tokio::io::{AsyncReadExt, AsyncSeekExt, SeekFrom}; - - let mut cloned = file.try_clone().await?; - cloned.seek(SeekFrom::Start(offset)).await?; - - let mut buffer = vec![0u8; size]; - cloned.read_exact(&mut buffer).await?; - - Ok(Self { - data: Bytes::from(buffer), - pos: 0, - }) - } - - /// Historical name for `from_file_read`. - #[deprecated( - since = "1.0.0-beta.8", - note = "use from_file_read; this method performs normal reads into owned Bytes" - )] - pub async fn from_file_mmap(file: &tokio::fs::File, offset: u64, size: usize) -> Result { - Self::from_file_read(file, offset, size).await - } - - /// Get the remaining data as Bytes (zero-copy). - /// - /// This returns a slice of the remaining data without copying. - /// The returned Bytes shares the underlying memory with this reader. - /// - /// # Example - /// - /// ```ignore - /// let remaining = reader.remaining_bytes(); - /// println!("Remaining: {} bytes", remaining.len()); - /// ``` - pub fn remaining_bytes(&self) -> Bytes { - self.data.slice(self.pos..) - } - - /// Get the total length of the data. - pub fn len(&self) -> usize { - self.data.len() - } - - /// Check if the reader has reached the end. - pub fn is_empty(&self) -> bool { - self.pos >= self.data.len() - } - - /// Get the current read position. - pub fn position(&self) -> usize { - self.pos - } -} - -impl AsyncRead for BytesBufferedReader { - fn poll_read(mut self: Pin<&mut Self>, _cx: &mut Context<'_>, buf: &mut ReadBuf<'_>) -> Poll> { - let remaining = self.data.len() - self.pos; - if remaining == 0 { - return Poll::Ready(Ok(())); - } - - let to_read = std::cmp::min(remaining, buf.remaining()); - let slice = &self.data[self.pos..self.pos + to_read]; - buf.put_slice(slice); - self.pos += to_read; - - Poll::Ready(Ok(())) - } -} - -impl std::fmt::Debug for BytesBufferedReader { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.debug_struct("BytesBufferedReader") - .field("data_len", &self.data.len()) - .field("pos", &self.pos) - .field("remaining", &(self.data.len() - self.pos)) - .finish() - } -} - -#[cfg(test)] -mod tests { - use super::*; - use std::path::PathBuf; - use tokio::io::AsyncReadExt; - - fn temp_file_path(test_name: &str) -> PathBuf { - let nonce = std::time::SystemTime::now() - .duration_since(std::time::UNIX_EPOCH) - .expect("system time should be after unix epoch") - .as_nanos(); - std::env::temp_dir().join(format!("rustfs-io-core-{test_name}-{}-{nonce}", std::process::id())) - } - - #[tokio::test] - async fn test_from_bytes() { - let data = Bytes::from("hello world"); - let mut reader = BytesBufferedReader::from_bytes(data.clone()); - - let mut buf = [0u8; 11]; - let n = reader.read(&mut buf[..]).await.unwrap(); - - assert_eq!(n, 11); - assert_eq!(&buf[..n], b"hello world"); - } - - #[tokio::test] - async fn test_preferred_reader_alias() { - let data = Bytes::from("hello world"); - let mut reader = BytesBufferedReader::from_bytes(data); - - let mut buf = [0u8; 5]; - let n = reader.read(&mut buf[..]).await.expect("read bytes from alias"); - - assert_eq!(n, 5); - assert_eq!(&buf[..n], b"hello"); - } - - #[tokio::test] - async fn test_from_file_read_reads_requested_range() { - let path = temp_file_path("from-file-read"); - tokio::fs::write(&path, b"hello world") - .await - .expect("write temp file for reader test"); - - let file = tokio::fs::File::open(&path).await.expect("open temp file for reader test"); - let mut reader = BytesBufferedReader::from_file_read(&file, 6, 5) - .await - .expect("read requested range into Bytes"); - - let mut output = Vec::new(); - reader.read_to_end(&mut output).await.expect("drain reader output"); - - assert_eq!(output, b"world"); - - let _ = tokio::fs::remove_file(path).await; - } - - #[tokio::test] - #[allow(deprecated)] - async fn test_from_file_mmap_legacy_alias_reads_requested_range() { - let path = temp_file_path("from-file-mmap"); - tokio::fs::write(&path, b"hello world") - .await - .expect("write temp file for legacy reader test"); - - let file = tokio::fs::File::open(&path) - .await - .expect("open temp file for legacy reader test"); - let mut reader = BytesBufferedReader::from_file_mmap(&file, 0, 5) - .await - .expect("read requested range through legacy alias"); - - let mut output = Vec::new(); - reader.read_to_end(&mut output).await.expect("drain legacy reader output"); - - assert_eq!(output, b"hello"); - - let _ = tokio::fs::remove_file(path).await; - } - - #[tokio::test] - async fn test_remaining_bytes() { - let data = Bytes::from("hello world"); - let reader = BytesBufferedReader::from_bytes(data); - - let remaining = reader.remaining_bytes(); - assert_eq!(remaining.len(), 11); - assert_eq!(&remaining[..], b"hello world"); - } - - #[tokio::test] - async fn test_position() { - let data = Bytes::from("hello world"); - let mut reader = BytesBufferedReader::from_bytes(data); - - assert_eq!(reader.position(), 0); - - let mut buf = [0u8; 5]; - reader.read_exact(&mut buf[..]).await.unwrap(); - - assert_eq!(reader.position(), 5); - } - - #[tokio::test] - async fn test_is_empty() { - let data = Bytes::from(""); - let reader = BytesBufferedReader::from_bytes(data); - assert!(reader.is_empty()); - - let data = Bytes::from("hello"); - let reader = BytesBufferedReader::from_bytes(data); - assert!(!reader.is_empty()); - } - - #[tokio::test] - #[allow(deprecated)] - async fn test_legacy_reader_alias() { - let data = Bytes::from("hello world"); - let mut reader = ZeroCopyObjectReader::from_bytes(data); - - let mut buf = [0u8; 5]; - let n = reader.read(&mut buf[..]).await.expect("read bytes through legacy alias"); - - assert_eq!(n, 5); - assert_eq!(&buf[..n], b"hello"); - } -} diff --git a/crates/io-core/src/scheduler.rs b/crates/io-core/src/scheduler.rs deleted file mode 100644 index 76afa58cf..000000000 --- a/crates/io-core/src/scheduler.rs +++ /dev/null @@ -1,882 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! I/O scheduler for adaptive buffer sizing and load management. -//! -//! This module provides the core I/O scheduling logic that determines -//! optimal buffer sizes, I/O strategies, and load management decisions. - -use crate::config::IoSchedulerConfig; -use crate::io_profile::{AccessPattern, StorageMedia, StorageProfile}; -use std::sync::atomic::{AtomicUsize, Ordering}; -use std::time::Duration; - -/// I/O priority levels. -#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Default)] -pub enum IoPriority { - /// High priority for small, latency-sensitive operations. - High, - /// Normal priority for standard operations. - #[default] - Normal, - /// Low priority for large, throughput-oriented operations. - Low, -} - -impl IoPriority { - /// Determine priority based on request size. - /// - /// A negative `size` means the size is unknown (-1 by convention) and maps - /// to `Normal`; casting it to `usize` would wrap to a huge value and - /// misclassify the request as `Low`. - pub fn from_size(size: i64, high_threshold: usize, low_threshold: usize) -> Self { - if size < 0 { - return IoPriority::Normal; - } - let size = size as usize; - if size < high_threshold { - IoPriority::High - } else if size > low_threshold { - IoPriority::Low - } else { - IoPriority::Normal - } - } - - /// Get the priority as a string for metrics labels. - pub fn as_str(&self) -> &'static str { - match self { - IoPriority::High => "high", - IoPriority::Normal => "normal", - IoPriority::Low => "low", - } - } - - /// Check if this is high priority. - pub fn is_high(&self) -> bool { - matches!(self, IoPriority::High) - } - - /// Check if this is normal priority. - pub fn is_normal(&self) -> bool { - matches!(self, IoPriority::Normal) - } - - /// Check if this is low priority. - pub fn is_low(&self) -> bool { - matches!(self, IoPriority::Low) - } -} - -impl std::fmt::Display for IoPriority { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - write!(f, "{}", self.as_str()) - } -} - -/// I/O load level. -#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Default)] -pub enum IoLoadLevel { - /// Low load - system is underutilized. - Low, - /// Medium load - system is moderately utilized. - #[default] - Medium, - /// High load - system is heavily utilized. - High, - /// Critical load - system is overloaded. - Critical, -} - -impl IoLoadLevel { - /// Get the load level as a string for metrics labels. - pub fn as_str(&self) -> &'static str { - match self { - IoLoadLevel::Low => "low", - IoLoadLevel::Medium => "medium", - IoLoadLevel::High => "high", - IoLoadLevel::Critical => "critical", - } - } - - /// Determine load level from wait time. - pub fn from_wait_time(wait_time: Duration, low_threshold: Duration, high_threshold: Duration) -> Self { - if wait_time <= low_threshold { - IoLoadLevel::Low - } else if wait_time <= high_threshold { - IoLoadLevel::Medium - } else if wait_time <= high_threshold * 2 { - IoLoadLevel::High - } else { - IoLoadLevel::Critical - } - } -} - -impl std::fmt::Display for IoLoadLevel { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - write!(f, "{}", self.as_str()) - } -} - -/// Bandwidth tier classification. -#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Default)] -pub enum BandwidthTier { - /// Low bandwidth (< 100 MB/s). - Low, - /// Medium bandwidth (100-500 MB/s). - #[default] - Medium, - /// High bandwidth (> 500 MB/s). - High, - /// Unknown bandwidth. - Unknown, -} - -impl BandwidthTier { - /// Determine bandwidth tier from bytes per second. - pub fn from_bps(bps: u64) -> Self { - const MB: u64 = 1024 * 1024; - if bps < 100 * MB { - BandwidthTier::Low - } else if bps < 500 * MB { - BandwidthTier::Medium - } else { - BandwidthTier::High - } - } - - /// Get the tier as a string for metrics labels. - pub fn as_str(&self) -> &'static str { - match self { - BandwidthTier::Low => "low", - BandwidthTier::Medium => "medium", - BandwidthTier::High => "high", - BandwidthTier::Unknown => "unknown", - } - } -} - -/// I/O strategy decision. -#[derive(Debug, Clone)] -pub struct IoStrategy { - /// Buffer size to use for I/O operations. - pub buffer_size: usize, - /// Buffer multiplier based on storage media. - pub buffer_multiplier: f64, - /// Whether to enable readahead. - pub enable_readahead: bool, - /// Whether to use buffered I/O. - pub use_buffered_io: bool, - - // Performance state - /// Current number of concurrent requests. - pub concurrent_requests: usize, - /// Observed bandwidth in bytes per second. - pub observed_bandwidth_bps: Option, - /// Bandwidth tier classification. - pub bandwidth_tier: BandwidthTier, - /// Current load level. - pub load_level: IoLoadLevel, - - // Priority - /// I/O priority for this operation. - pub priority: IoPriority, - - // Decision flags - /// Whether to throttle random I/O. - pub should_throttle_random_io: bool, - /// Whether to expand buffer for sequential access. - pub should_expand_for_sequential: bool, - /// Whether to reduce buffer due to concurrency. - pub should_reduce_for_concurrency: bool, - /// Whether to reduce buffer due to low bandwidth. - pub should_reduce_for_bandwidth: bool, -} - -impl Default for IoStrategy { - fn default() -> Self { - Self { - buffer_size: 128 * 1024, - buffer_multiplier: 1.0, - enable_readahead: true, - use_buffered_io: true, - concurrent_requests: 0, - observed_bandwidth_bps: None, - bandwidth_tier: BandwidthTier::Medium, - load_level: IoLoadLevel::Low, - priority: IoPriority::Normal, - should_throttle_random_io: false, - should_expand_for_sequential: false, - should_reduce_for_concurrency: false, - should_reduce_for_bandwidth: false, - } - } -} - -impl IoStrategy { - /// Create a new strategy with default values. - pub fn new() -> Self { - Self::default() - } - - /// Create a strategy for sequential access. - pub fn sequential(buffer_size: usize) -> Self { - Self { - buffer_size, - enable_readahead: true, - should_expand_for_sequential: true, - ..Self::default() - } - } - - /// Create a strategy for random access. - pub fn random(buffer_size: usize) -> Self { - Self { - buffer_size, - enable_readahead: false, - should_throttle_random_io: true, - ..Self::default() - } - } -} - -/// I/O load metrics. -#[derive(Debug, Clone, Default)] -pub struct IoLoadMetrics { - /// Number of samples in the current window. - pub sample_count: usize, - /// Total wait time in the window. - pub total_wait_time: Duration, - /// Maximum wait time in the window. - pub max_wait_time: Duration, - /// Average wait time. - pub avg_wait_time: Duration, - /// Current load level. - pub load_level: IoLoadLevel, -} - -impl IoLoadMetrics { - /// Create new load metrics. - pub fn new() -> Self { - Self::default() - } - - /// Add a wait time sample. - pub fn add_sample(&mut self, wait_time: Duration) { - self.sample_count += 1; - self.total_wait_time += wait_time; - if wait_time > self.max_wait_time { - self.max_wait_time = wait_time; - } - self.avg_wait_time = if self.sample_count > 0 { - self.total_wait_time / self.sample_count as u32 - } else { - Duration::ZERO - }; - } - - /// Update load level based on thresholds. - pub fn update_load_level(&mut self, low_threshold: Duration, high_threshold: Duration) { - self.load_level = IoLoadLevel::from_wait_time(self.avg_wait_time, low_threshold, high_threshold); - } - - /// Reset the metrics. - pub fn reset(&mut self) { - *self = Self::default(); - } -} - -/// I/O scheduler. -pub struct IoScheduler { - /// Scheduler configuration. - config: IoSchedulerConfig, - /// Active request counter. - active_requests: AtomicUsize, - /// Load metrics. - load_metrics: std::sync::Mutex, -} - -impl IoScheduler { - /// Create a new I/O scheduler with the given configuration. - pub fn new(config: IoSchedulerConfig) -> Self { - Self { - config, - active_requests: AtomicUsize::new(0), - load_metrics: std::sync::Mutex::new(IoLoadMetrics::new()), - } - } - - /// Create a new I/O scheduler with default configuration. - pub fn with_defaults() -> Self { - Self::new(IoSchedulerConfig::default()) - } - - /// Get the scheduler configuration. - pub fn config(&self) -> &IoSchedulerConfig { - &self.config - } - - /// Get the current number of active requests. - pub fn active_requests(&self) -> usize { - self.active_requests.load(Ordering::Relaxed) - } - - /// Increment the active request count. - pub fn increment_requests(&self) { - self.active_requests.fetch_add(1, Ordering::Relaxed); - } - - /// Decrement the active request count. - pub fn decrement_requests(&self) { - self.active_requests.fetch_sub(1, Ordering::Relaxed); - } - - /// Calculate I/O strategy for a request. - pub fn calculate_strategy(&self, file_size: i64, permit_wait_time: Duration, is_sequential: bool) -> IoStrategy { - let concurrent_requests = self.active_requests.load(Ordering::Relaxed); - - // Determine priority based on file size - let priority = IoPriority::from_size( - file_size, - self.config.high_priority_size_threshold, - self.config.low_priority_size_threshold, - ); - - // Determine load level - let load_level = - IoLoadLevel::from_wait_time(permit_wait_time, self.config.load_low_threshold(), self.config.load_high_threshold()); - - // Calculate base buffer size - let base_buffer = self.config.base_buffer_size; - - // Adjust for concurrency - let concurrency_factor = match concurrent_requests { - 0..=2 => 1.0, - 3..=4 => 0.75, - 5..=8 => 0.5, - _ => 0.4, - }; - - // Adjust for load level - let load_factor = match load_level { - IoLoadLevel::Low => 1.2, - IoLoadLevel::Medium => 1.0, - IoLoadLevel::High => 0.7, - IoLoadLevel::Critical => 0.5, - }; - - // Adjust for access pattern - let sequential_factor = if is_sequential { 1.5 } else { 1.0 }; - - // Calculate final buffer size - let buffer_size = (base_buffer as f64 * concurrency_factor * load_factor * sequential_factor) as usize; - let buffer_size = buffer_size.clamp(self.config.min_buffer_size, self.config.max_buffer_size); - - IoStrategy { - buffer_size, - buffer_multiplier: concurrency_factor * load_factor * sequential_factor, - enable_readahead: is_sequential && load_level != IoLoadLevel::Critical, - use_buffered_io: true, - concurrent_requests, - observed_bandwidth_bps: None, - bandwidth_tier: BandwidthTier::Unknown, - load_level, - priority, - should_throttle_random_io: !is_sequential && load_level >= IoLoadLevel::High, - should_expand_for_sequential: is_sequential && load_level <= IoLoadLevel::Medium, - should_reduce_for_concurrency: concurrent_requests > 4, - should_reduce_for_bandwidth: false, - } - } - - /// Calculate multi-factor I/O strategy. - pub fn calculate_multi_factor_strategy( - &self, - file_size: i64, - permit_wait_time: Duration, - is_sequential: bool, - storage_profile: Option<&StorageProfile>, - ) -> IoStrategy { - let mut strategy = self.calculate_strategy(file_size, permit_wait_time, is_sequential); - - // Apply storage profile adjustments - if let Some(profile) = storage_profile { - // Adjust buffer size based on storage media - let media_factor = match profile.media { - StorageMedia::Nvme => 1.5, - StorageMedia::Ssd => 1.2, - StorageMedia::Hdd => 0.8, - StorageMedia::Unknown => 1.0, - }; - - strategy.buffer_size = (strategy.buffer_size as f64 * media_factor).min(self.config.max_buffer_size as f64) as usize; - - // Apply sequential boost if applicable - if is_sequential { - strategy.buffer_size = (strategy.buffer_size as f64 * profile.sequential_boost_multiplier) - .min(self.config.max_buffer_size as f64) as usize; - } - - // Apply random penalty if applicable - if !is_sequential { - strategy.buffer_size = (strategy.buffer_size as f64 * profile.random_penalty_multiplier) - .max(self.config.min_buffer_size as f64) as usize; - } - - // Update readahead preference - strategy.enable_readahead = strategy.enable_readahead && profile.prefers_readahead; - } - - strategy - } - - /// Record a wait time sample for load tracking. - pub fn record_wait_time(&self, wait_time: Duration) { - if let Ok(mut metrics) = self.load_metrics.lock() { - metrics.add_sample(wait_time); - metrics.update_load_level(self.config.load_low_threshold(), self.config.load_high_threshold()); - } - } - - /// Get current load metrics. - pub fn load_metrics(&self) -> IoLoadMetrics { - if let Ok(metrics) = self.load_metrics.lock() { - metrics.clone() - } else { - IoLoadMetrics::default() - } - } -} - -impl Default for IoScheduler { - fn default() -> Self { - Self::with_defaults() - } -} - -// ============================================================================ -// Buffer Size Calculation Functions -// ============================================================================ - -/// Constants for buffer size calculations. -pub const KI_B: usize = 1024; -pub const MI_B: usize = 1024 * 1024; - -/// Get concurrency-aware buffer size. -/// -/// Adjusts buffer size based on the current level of concurrent requests. -/// Higher concurrency leads to smaller buffers to reduce memory pressure. -/// -/// # Arguments -/// -/// * `file_size` - Size of the file being read (-1 if unknown) -/// * `base_buffer_size` - Base buffer size from workload profile -/// -/// # Returns -/// -/// Adjusted buffer size in bytes -pub fn get_concurrency_aware_buffer_size(file_size: i64, base_buffer_size: usize) -> usize { - // Get current concurrency level from global counter - let concurrent_requests = 1; // Default to 1 if no global counter available - - // Define concurrency thresholds - let medium_threshold = 4; - let high_threshold = 8; - - // Calculate adaptive multiplier based on concurrency - let adaptive_multiplier = if concurrent_requests <= 2 { - // Low concurrency (1-2): use full buffer size - 1.0 - } else if concurrent_requests <= medium_threshold { - // Medium concurrency (3-4): slightly reduce buffer size (75% of base) - 0.75 - } else if concurrent_requests <= high_threshold { - // Higher concurrency (5-8): more aggressive reduction (50% of base) - 0.5 - } else { - // Very high concurrency (>8): minimize memory per request (40% of base) - 0.4 - }; - - // Calculate the adjusted buffer size - let adjusted_size = (base_buffer_size as f64 * adaptive_multiplier) as usize; - - // Ensure we stay within reasonable bounds - let min_buffer = if file_size > 0 && file_size < 100 * KI_B as i64 { - 32 * KI_B // For very small files, use minimum buffer - } else { - 64 * KI_B // Standard minimum buffer size - }; - - let max_buffer = if concurrent_requests > high_threshold { - 256 * KI_B // Cap at 256KB for high concurrency - } else { - MI_B // Cap at 1MB for lower concurrency - }; - - adjusted_size.clamp(min_buffer, max_buffer) -} - -/// Advanced concurrency-aware buffer sizing with file size optimization. -/// -/// This enhanced version considers both concurrency level and file size patterns -/// to provide even better performance characteristics. -/// -/// # Arguments -/// -/// * `file_size` - Size of the file being read (-1 if unknown) -/// * `base_buffer_size` - Baseline buffer size from workload profile -/// * `is_sequential` - Whether this is a sequential read (hint for optimization) -/// * `concurrent_requests` - Current number of concurrent requests -/// -/// # Returns -/// -/// Optimized buffer size in bytes -pub fn get_advanced_buffer_size( - file_size: i64, - base_buffer_size: usize, - is_sequential: bool, - concurrent_requests: usize, -) -> usize { - // For very small files, use smaller buffers regardless of concurrency - if file_size > 0 && file_size < 256 * KI_B as i64 { - return (file_size as usize / 4).clamp(16 * KI_B, 64 * KI_B); - } - - // Base calculation from standard function - let standard_size = get_concurrency_aware_buffer_size(file_size, base_buffer_size); - - let medium_threshold = 4; - let high_threshold = 8; - - // For sequential reads, we can be more aggressive with buffer sizes - if is_sequential && concurrent_requests <= medium_threshold { - // Boost buffer size for sequential reads under low concurrency - let boosted = (standard_size as f64 * 1.5) as usize; - return boosted.min(MI_B); - } - - // For random reads under high concurrency, reduce buffer size - if !is_sequential && concurrent_requests > high_threshold { - let reduced = (standard_size as f64 * 0.7) as usize; - return reduced.max(32 * KI_B); - } - - standard_size -} - -/// Get buffer size with storage media optimization. -/// -/// Adjusts buffer size based on storage media characteristics. -/// -/// # Arguments -/// -/// * `base_size` - Base buffer size -/// * `media` - Storage media type -/// -/// # Returns -/// -/// Optimized buffer size for the storage media -pub fn get_buffer_size_for_media(base_size: usize, media: StorageMedia) -> usize { - let multiplier = match media { - StorageMedia::Nvme => 1.5, // NVMe can handle larger buffers - StorageMedia::Ssd => 1.2, // SSD benefits from moderate buffers - StorageMedia::Hdd => 0.8, // HDD prefers smaller buffers to reduce seek overhead - StorageMedia::Unknown => 1.0, - }; - - (base_size as f64 * multiplier).min(MI_B as f64) as usize -} - -/// Calculate optimal buffer size using multi-factor analysis. -/// -/// This is the main entry point for buffer size calculation, considering -/// all factors: concurrency, storage media, access pattern, and load. -/// -/// # Arguments -/// -/// * `file_size` - Size of the file being read -/// * `base_buffer_size` - Base buffer size -/// * `is_sequential` - Whether access is sequential -/// * `concurrent_requests` - Current concurrency level -/// * `media` - Storage media type -/// * `load_level` - Current I/O load level -/// -/// # Returns -/// -/// Optimally calculated buffer size -pub fn calculate_optimal_buffer_size( - file_size: i64, - base_buffer_size: usize, - is_sequential: bool, - concurrent_requests: usize, - media: StorageMedia, - load_level: IoLoadLevel, -) -> usize { - // Start with advanced buffer size calculation - let mut buffer_size = get_advanced_buffer_size(file_size, base_buffer_size, is_sequential, concurrent_requests); - - // Apply storage media optimization - buffer_size = get_buffer_size_for_media(buffer_size, media); - - // Apply load-based adjustment - let load_multiplier = match load_level { - IoLoadLevel::Low => 1.2, - IoLoadLevel::Medium => 1.0, - IoLoadLevel::High => 0.7, - IoLoadLevel::Critical => 0.5, - }; - - buffer_size = (buffer_size as f64 * load_multiplier) as usize; - - // Final bounds check - buffer_size.clamp(32 * KI_B, MI_B) -} - -/// I/O scheduling context for multi-factor strategy calculation. -#[derive(Debug, Clone)] -pub struct IoSchedulingContext { - /// File size in bytes (-1 if unknown). - pub file_size: i64, - /// Base buffer size from configuration. - pub base_buffer_size: usize, - /// Time spent waiting for permit. - pub permit_wait_duration: Duration, - /// Whether access is sequential. - pub is_sequential_hint: bool, - /// Detected access pattern. - pub access_pattern: AccessPattern, - /// Detected storage media. - pub storage_media: StorageMedia, - /// Observed bandwidth in bytes per second. - pub observed_bandwidth_bps: Option, - /// Current concurrent request count. - pub concurrent_requests: usize, -} - -impl Default for IoSchedulingContext { - fn default() -> Self { - Self { - file_size: -1, - base_buffer_size: 128 * KI_B, - permit_wait_duration: Duration::ZERO, - is_sequential_hint: true, - access_pattern: AccessPattern::Unknown, - storage_media: StorageMedia::Unknown, - observed_bandwidth_bps: None, - concurrent_requests: 1, - } - } -} - -impl IoSchedulingContext { - /// Create a new scheduling context. - pub fn new(file_size: i64, base_buffer_size: usize) -> Self { - Self { - file_size, - base_buffer_size, - ..Self::default() - } - } - - /// Builder pattern: set sequential hint. - pub fn with_sequential(mut self, is_sequential: bool) -> Self { - self.is_sequential_hint = is_sequential; - self.access_pattern = if is_sequential { - AccessPattern::Sequential - } else { - AccessPattern::Random - }; - self - } - - /// Builder pattern: set storage media. - pub fn with_media(mut self, media: StorageMedia) -> Self { - self.storage_media = media; - self - } - - /// Builder pattern: set bandwidth. - pub fn with_bandwidth(mut self, bps: u64) -> Self { - self.observed_bandwidth_bps = Some(bps); - self - } - - /// Builder pattern: set concurrency. - pub fn with_concurrency(mut self, count: usize) -> Self { - self.concurrent_requests = count; - self - } - - /// Builder pattern: set wait duration. - pub fn with_wait_duration(mut self, duration: Duration) -> Self { - self.permit_wait_duration = duration; - self - } -} - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn test_io_priority() { - assert_eq!(IoPriority::from_size(1024, 64 * 1024, 4 * 1024 * 1024), IoPriority::High); - assert_eq!(IoPriority::from_size(1024 * 1024, 64 * 1024, 4 * 1024 * 1024), IoPriority::Normal); - assert_eq!(IoPriority::from_size(10 * 1024 * 1024, 64 * 1024, 4 * 1024 * 1024), IoPriority::Low); - } - - #[test] - fn test_io_priority_unknown_size_is_normal() { - // -1 means "size unknown" and must not wrap to usize::MAX (=> Low). - assert_eq!(IoPriority::from_size(-1, 64 * 1024, 4 * 1024 * 1024), IoPriority::Normal); - assert_eq!(IoPriority::from_size(i64::MIN, 64 * 1024, 4 * 1024 * 1024), IoPriority::Normal); - } - - #[test] - fn test_io_load_level() { - let low = Duration::from_millis(5); - let high = Duration::from_millis(50); - - assert_eq!(IoLoadLevel::from_wait_time(Duration::from_millis(1), low, high), IoLoadLevel::Low); - assert_eq!(IoLoadLevel::from_wait_time(Duration::from_millis(20), low, high), IoLoadLevel::Medium); - assert_eq!(IoLoadLevel::from_wait_time(Duration::from_millis(60), low, high), IoLoadLevel::High); - assert_eq!(IoLoadLevel::from_wait_time(Duration::from_millis(150), low, high), IoLoadLevel::Critical); - } - - #[test] - fn test_bandwidth_tier() { - assert_eq!(BandwidthTier::from_bps(50 * 1024 * 1024), BandwidthTier::Low); - assert_eq!(BandwidthTier::from_bps(200 * 1024 * 1024), BandwidthTier::Medium); - assert_eq!(BandwidthTier::from_bps(600 * 1024 * 1024), BandwidthTier::High); - } - - #[test] - fn test_io_strategy_default() { - let strategy = IoStrategy::default(); - assert!(strategy.buffer_size > 0); - assert!(strategy.enable_readahead); - } - - #[test] - fn test_io_scheduler() { - let scheduler = IoScheduler::with_defaults(); - - let strategy = scheduler.calculate_strategy(1024 * 1024, Duration::from_millis(5), true); - assert!(strategy.buffer_size > 0); - assert!(strategy.enable_readahead); - assert_eq!(strategy.load_level, IoLoadLevel::Low); - } - - #[test] - fn test_io_scheduler_with_concurrency() { - let scheduler = IoScheduler::with_defaults(); - - // Simulate concurrent requests - scheduler.increment_requests(); - scheduler.increment_requests(); - scheduler.increment_requests(); - - let strategy = scheduler.calculate_strategy(1024 * 1024, Duration::from_millis(5), true); - assert_eq!(strategy.concurrent_requests, 3); - } - - #[test] - fn test_load_metrics() { - let mut metrics = IoLoadMetrics::new(); - - metrics.add_sample(Duration::from_millis(10)); - metrics.add_sample(Duration::from_millis(20)); - metrics.add_sample(Duration::from_millis(30)); - - assert_eq!(metrics.sample_count, 3); - assert_eq!(metrics.avg_wait_time, Duration::from_millis(20)); - assert_eq!(metrics.max_wait_time, Duration::from_millis(30)); - } - - #[test] - fn test_get_concurrency_aware_buffer_size() { - // Test with default concurrency (1) - let size = get_concurrency_aware_buffer_size(1024 * 1024, 128 * KI_B); - assert!(size >= 64 * KI_B); - assert!(size <= MI_B); - - // Test with small file - let size = get_concurrency_aware_buffer_size(50 * KI_B as i64, 128 * KI_B); - assert!(size >= 32 * KI_B); - } - - #[test] - fn test_get_advanced_buffer_size() { - // Sequential read with low concurrency - let size = get_advanced_buffer_size(10 * MI_B as i64, 128 * KI_B, true, 2); - assert!(size >= 128 * KI_B); - - // Random read with high concurrency - let size = get_advanced_buffer_size(10 * MI_B as i64, 128 * KI_B, false, 10); - assert!(size >= 32 * KI_B); - - // Very small file - let size = get_advanced_buffer_size(100 * KI_B as i64, 128 * KI_B, true, 1); - assert!(size <= 64 * KI_B); - } - - #[test] - fn test_get_buffer_size_for_media() { - let base = 128 * KI_B; - - // NVMe should get larger buffers - let nvme_size = get_buffer_size_for_media(base, StorageMedia::Nvme); - assert!(nvme_size > base); - - // SSD should get slightly larger buffers - let ssd_size = get_buffer_size_for_media(base, StorageMedia::Ssd); - assert!(ssd_size > base); - - // HDD should get smaller buffers - let hdd_size = get_buffer_size_for_media(base, StorageMedia::Hdd); - assert!(hdd_size < base); - } - - #[test] - fn test_calculate_optimal_buffer_size() { - // Low load, sequential, NVMe - let size = calculate_optimal_buffer_size(10 * MI_B as i64, 128 * KI_B, true, 2, StorageMedia::Nvme, IoLoadLevel::Low); - assert!(size >= 32 * KI_B); - assert!(size <= MI_B); - - // Critical load, random, HDD - let size = - calculate_optimal_buffer_size(10 * MI_B as i64, 128 * KI_B, false, 10, StorageMedia::Hdd, IoLoadLevel::Critical); - assert!(size >= 32 * KI_B); - assert!(size <= MI_B); - } - - #[test] - fn test_io_scheduling_context() { - let ctx = IoSchedulingContext::new(10 * MI_B as i64, 256 * KI_B) - .with_sequential(true) - .with_media(StorageMedia::Nvme) - .with_bandwidth(500 * MI_B as u64) - .with_concurrency(4); - - assert_eq!(ctx.file_size, 10 * MI_B as i64); - assert_eq!(ctx.base_buffer_size, 256 * KI_B); - assert!(ctx.is_sequential_hint); - assert_eq!(ctx.storage_media, StorageMedia::Nvme); - assert_eq!(ctx.observed_bandwidth_bps, Some(500 * MI_B as u64)); - assert_eq!(ctx.concurrent_requests, 4); - } -} diff --git a/crates/io-core/src/shared_memory.rs b/crates/io-core/src/shared_memory.rs deleted file mode 100644 index e21781beb..000000000 --- a/crates/io-core/src/shared_memory.rs +++ /dev/null @@ -1,320 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Shared memory pool for zero-copy data sharing. -//! -//! This module provides Arc-based shared memory management for -//! efficient cross-task data passing without serialization. - -use std::convert::AsRef; -use std::ops::Deref; -use std::sync::Arc; -use std::sync::atomic::{AtomicU64, Ordering}; -use std::time::Instant; - -/// Shared memory pool configuration. -#[derive(Debug, Clone)] -pub struct SharedMemoryConfig { - /// Whether shared memory is enabled - pub enabled: bool, - - /// Maximum pool size in bytes - pub max_pool_size: usize, - - /// Maximum object size in bytes - pub max_object_size: usize, -} - -impl Default for SharedMemoryConfig { - fn default() -> Self { - Self { - enabled: true, - max_pool_size: 100 * 1024 * 1024, // 100MB - max_object_size: 10 * 1024 * 1024, // 10MB - } - } -} - -/// Shared memory pool statistics. -#[derive(Debug, Default)] -pub struct SharedMemoryStats { - /// Total number of objects created - pub total_objects: AtomicU64, - - /// Total number of shared references - pub total_shared_refs: AtomicU64, - - /// Current memory usage in bytes - pub current_memory: AtomicU64, - - /// Peak memory usage in bytes - pub peak_memory: AtomicU64, -} - -/// Arc data metadata. -#[derive(Clone, Debug)] -pub struct ArcMetadata { - /// Size of the data (if measurable) - pub size: Option, - - /// Creation timestamp - pub created_at: Instant, -} - -/// Arc-based data wrapper for zero-copy sharing. -/// -/// This wrapper uses Arc to enable shared ownership of data -/// across multiple tasks without copying. -pub struct ArcData { - /// The wrapped data - inner: Arc, - - /// Metadata about the data - metadata: ArcMetadata, -} - -impl Clone for ArcData { - fn clone(&self) -> Self { - Self { - inner: Arc::clone(&self.inner), - metadata: self.metadata.clone(), - } - } -} - -impl ArcData { - /// Create a new ArcData wrapper. - pub fn new(data: T) -> Self { - ArcData { - inner: Arc::new(data), - metadata: ArcMetadata { - size: None, - created_at: Instant::now(), - }, - } - } - - /// Create a new ArcData wrapper with known size. - pub fn with_size(data: T, size: usize) -> Self { - ArcData { - inner: Arc::new(data), - metadata: ArcMetadata { - size: Some(size), - created_at: Instant::now(), - }, - } - } - - /// Get the reference count. - pub fn ref_count(&self) -> usize { - Arc::strong_count(&self.inner) - } - - /// Convert into the underlying Arc. - pub fn into_arc(self) -> Arc { - self.inner - } - - /// Get the metadata. - pub fn metadata(&self) -> &ArcMetadata { - &self.metadata - } - - /// Get the size if known. - pub fn size(&self) -> Option { - self.metadata.size - } -} - -impl AsRef for ArcData { - fn as_ref(&self) -> &T { - &self.inner - } -} - -impl Deref for ArcData { - type Target = T; - - fn deref(&self) -> &Self::Target { - &self.inner - } -} - -impl std::fmt::Debug for ArcData { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.debug_struct("ArcData") - .field("ref_count", &self.ref_count()) - .field("metadata", &self.metadata) - .finish() - } -} - -/// Shared memory pool for managing Arc-based shared data. -pub struct SharedMemoryPool { - config: SharedMemoryConfig, - stats: SharedMemoryStats, -} - -impl SharedMemoryPool { - /// Create a new shared memory pool with the given configuration. - pub fn new(config: SharedMemoryConfig) -> Self { - Self { - config, - stats: SharedMemoryStats::default(), - } - } - - /// Create a new shared memory pool with default configuration. - pub fn with_defaults() -> Self { - Self::new(SharedMemoryConfig::default()) - } - - /// Create shared data. - /// - /// This method wraps the data in an ArcData for zero-copy sharing. - pub fn create(&self, data: T) -> ArcData { - self.stats.total_objects.fetch_add(1, Ordering::Relaxed); - ArcData::new(data) - } - - /// Create shared data with known size. - /// - /// This method tracks memory usage for statistics. - pub fn create_with_size(&self, data: T, size: usize) -> ArcData { - self.stats.total_objects.fetch_add(1, Ordering::Relaxed); - - // Update memory statistics - self.stats.current_memory.fetch_add(size as u64, Ordering::Relaxed); - - // Update peak memory - let current = self.stats.current_memory.load(Ordering::Relaxed); - let mut peak = self.stats.peak_memory.load(Ordering::Relaxed); - if current > peak { - peak = current; - self.stats.peak_memory.store(peak, Ordering::Relaxed); - } - - ArcData::with_size(data, size) - } - - /// Share data by increasing reference count. - /// - /// This method creates a new ArcData that shares the underlying data - /// without copying. - pub fn share(&self, data: &ArcData) -> ArcData { - self.stats.total_shared_refs.fetch_add(1, Ordering::Relaxed); - data.clone() - } - - /// Get the statistics for this pool. - pub fn stats(&self) -> &SharedMemoryStats { - &self.stats - } - - /// Get the configuration for this pool. - pub fn config(&self) -> &SharedMemoryConfig { - &self.config - } - - /// Check if the pool is enabled. - pub fn is_enabled(&self) -> bool { - self.config.enabled - } -} - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn test_arc_data_new() { - let data = vec![1u8, 2, 3, 4, 5]; - let arc_data = ArcData::new(data.clone()); - - assert_eq!(arc_data.as_ref(), &data); - assert_eq!(arc_data.ref_count(), 1); - } - - #[test] - fn test_arc_data_clone() { - let data = vec![1u8, 2, 3, 4, 5]; - let arc_data = ArcData::new(data); - - assert_eq!(arc_data.ref_count(), 1); - - let arc_data2 = arc_data.clone(); - assert_eq!(arc_data.ref_count(), 2); - assert_eq!(arc_data2.ref_count(), 2); - - let arc_data3 = arc_data.clone(); - assert_eq!(arc_data.ref_count(), 3); - assert_eq!(arc_data2.ref_count(), 3); - assert_eq!(arc_data3.ref_count(), 3); - } - - #[test] - fn test_arc_data_deref() { - let data = vec![1u8, 2, 3, 4, 5]; - let arc_data = ArcData::new(data); - - // Test Deref trait - assert_eq!(arc_data.len(), 5); - assert_eq!(arc_data[0], 1); - } - - #[test] - fn test_shared_memory_pool_create() { - let pool = SharedMemoryPool::with_defaults(); - let data = vec![1u8, 2, 3, 4, 5]; - - let arc_data = pool.create(data.clone()); - - assert_eq!(arc_data.as_ref(), &data); - assert_eq!(pool.stats().total_objects.load(Ordering::Relaxed), 1); - } - - #[test] - fn test_shared_memory_pool_share() { - let pool = SharedMemoryPool::with_defaults(); - let data = vec![1u8, 2, 3, 4, 5]; - - let arc_data = pool.create(data); - assert_eq!(arc_data.ref_count(), 1); - - let shared = pool.share(&arc_data); - assert_eq!(arc_data.ref_count(), 2); - assert_eq!(shared.ref_count(), 2); - assert_eq!(pool.stats().total_shared_refs.load(Ordering::Relaxed), 1); - } - - #[test] - fn test_shared_memory_pool_with_size() { - let pool = SharedMemoryPool::with_defaults(); - let data = vec![1u8; 1024]; - - let arc_data = pool.create_with_size(data, 1024); - - assert_eq!(arc_data.size(), Some(1024)); - assert_eq!(pool.stats().current_memory.load(Ordering::Relaxed), 1024); - } - - #[test] - fn test_default_config() { - let config = SharedMemoryConfig::default(); - - assert!(config.enabled); - assert_eq!(config.max_pool_size, 100 * 1024 * 1024); - assert_eq!(config.max_object_size, 10 * 1024 * 1024); - } -} diff --git a/crates/io-core/src/timeout_wrapper.rs b/crates/io-core/src/timeout_wrapper.rs deleted file mode 100644 index 9d9ce4b22..000000000 --- a/crates/io-core/src/timeout_wrapper.rs +++ /dev/null @@ -1,501 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! Timeout wrapper for I/O operations. -//! -//! This module provides timeout management for I/O operations with -//! dynamic timeout calculation based on operation size. - -use std::sync::atomic::{AtomicU64, Ordering}; -use std::time::{Duration, Instant}; - -/// Timeout configuration. -#[derive(Debug, Clone)] -pub struct TimeoutConfig { - /// Base timeout for small operations. - pub base_timeout: Duration, - /// Timeout per MB of data. - pub timeout_per_mb: Duration, - /// Maximum timeout. - pub max_timeout: Duration, - /// Minimum timeout. - pub min_timeout: Duration, - /// GetObject operation timeout. - pub get_object_timeout: Duration, - /// PutObject operation timeout. - pub put_object_timeout: Duration, - /// ListObjects operation timeout. - pub list_objects_timeout: Duration, - /// Whether dynamic timeout is enabled. - pub enable_dynamic_timeout: bool, -} - -impl Default for TimeoutConfig { - fn default() -> Self { - Self { - base_timeout: Duration::from_secs(5), - timeout_per_mb: Duration::from_millis(100), - max_timeout: Duration::from_secs(300), - min_timeout: Duration::from_secs(1), - get_object_timeout: Duration::from_secs(30), - put_object_timeout: Duration::from_secs(60), - list_objects_timeout: Duration::from_secs(10), - enable_dynamic_timeout: true, - } - } -} - -impl TimeoutConfig { - /// Create new timeout configuration. - pub fn new() -> Self { - Self::default() - } - - /// Calculate dynamic timeout based on size. - pub fn calculate_timeout(&self, size_bytes: u64) -> Duration { - if !self.enable_dynamic_timeout { - return self.base_timeout; - } - - let mb = size_bytes as f64 / (1024.0 * 1024.0); - let timeout = self.base_timeout + self.timeout_per_mb.mul_f64(mb); - timeout.clamp(self.min_timeout, self.max_timeout) - } - - /// Validate the configuration. - pub fn validate(&self) -> Result<(), TimeoutError> { - if self.min_timeout > self.max_timeout { - return Err(TimeoutError::InvalidConfig("min_timeout must be <= max_timeout".to_string())); - } - if self.base_timeout < self.min_timeout || self.base_timeout > self.max_timeout { - return Err(TimeoutError::InvalidConfig( - "base_timeout must be between min_timeout and max_timeout".to_string(), - )); - } - Ok(()) - } -} - -/// Timeout error. -#[derive(Debug, Clone, thiserror::Error)] -pub enum TimeoutError { - /// Operation timed out. - #[error("Operation timed out after {0:?}")] - TimedOut(Duration), - /// Invalid configuration. - #[error("Invalid timeout config: {0}")] - InvalidConfig(String), -} - -/// Operation progress tracker. -#[derive(Debug)] -pub struct OperationProgress { - /// Total size (if known). - pub total_size: Option, - /// Bytes processed. - bytes_processed: AtomicU64, - /// Last update time. - last_update: std::sync::Mutex, - /// Stale timeout. - stale_timeout: Duration, - /// Start time for transfer rate calculation. - start_time: Instant, -} - -impl OperationProgress { - /// Create new operation progress. - pub fn new(total_size: Option, stale_timeout: Duration) -> Self { - Self { - total_size, - bytes_processed: AtomicU64::new(0), - last_update: std::sync::Mutex::new(Instant::now()), - stale_timeout, - start_time: Instant::now(), - } - } - - /// Update progress. - pub fn update(&self, bytes: u64) { - self.bytes_processed.store(bytes, Ordering::Relaxed); - if let Ok(mut last) = self.last_update.lock() { - *last = Instant::now(); - } - } - - /// Add to progress. - pub fn add(&self, bytes: u64) { - self.bytes_processed.fetch_add(bytes, Ordering::Relaxed); - if let Ok(mut last) = self.last_update.lock() { - *last = Instant::now(); - } - } - - /// Get current progress. - pub fn current(&self) -> u64 { - self.bytes_processed.load(Ordering::Relaxed) - } - - /// Check if progress is stale. - pub fn is_stale(&self) -> bool { - if let Ok(last) = self.last_update.lock() { - last.elapsed() > self.stale_timeout - } else { - false - } - } - - /// Get progress percentage. - pub fn progress_percent(&self) -> Option { - self.total_size.map(|total| { - if total == 0 { - 100.0 - } else { - let processed = self.bytes_processed.load(Ordering::Relaxed); - (processed as f64 / total as f64 * 100.0).min(100.0) - } - }) - } - - /// Get remaining bytes. - pub fn remaining(&self) -> Option { - self.total_size.map(|total| { - let processed = self.bytes_processed.load(Ordering::Relaxed); - total.saturating_sub(processed) - }) - } - - /// Calculate transfer rate in bytes per second. - /// - /// Returns 0 if no time has elapsed or no data transferred. - pub fn transfer_rate(&self) -> u64 { - let processed = self.bytes_processed.load(Ordering::Relaxed); - if processed == 0 { - return 0; - } - - let elapsed = self.start_time.elapsed().as_secs_f64(); - if elapsed > 0.0 { - (processed as f64 / elapsed) as u64 - } else { - 0 - } - } -} - -/// Request timeout wrapper. -pub struct RequestTimeoutWrapper { - /// Configuration. - config: TimeoutConfig, - /// Start time. - start_time: Instant, - /// Operation progress. - progress: Option, -} - -impl RequestTimeoutWrapper { - /// Create a new timeout wrapper. - pub fn new(config: TimeoutConfig) -> Self { - Self { - config, - start_time: Instant::now(), - progress: None, - } - } - - /// Create with progress tracking. - pub fn with_progress(config: TimeoutConfig, total_size: Option, stale_timeout: Duration) -> Self { - Self { - config, - start_time: Instant::now(), - progress: Some(OperationProgress::new(total_size, stale_timeout)), - } - } - - /// Get the configuration. - pub fn config(&self) -> &TimeoutConfig { - &self.config - } - - /// Get elapsed time. - pub fn elapsed(&self) -> Duration { - self.start_time.elapsed() - } - - /// Get remaining time. - pub fn remaining(&self, timeout: Duration) -> Option { - let elapsed = self.elapsed(); - if elapsed >= timeout { None } else { Some(timeout - elapsed) } - } - - /// Check if timed out. - pub fn is_timed_out(&self, size: Option) -> bool { - let timeout = self.get_timeout(size); - self.elapsed() > timeout - } - - /// Get the timeout for a given size. - pub fn get_timeout(&self, size: Option) -> Duration { - if self.config.enable_dynamic_timeout { - if let Some(s) = size { - self.config.calculate_timeout(s) - } else { - self.config.base_timeout - } - } else { - self.config.base_timeout - } - } - - /// Check if timed out and return error if so. - pub fn check_timeout(&self, size: Option) -> Result<(), TimeoutError> { - if self.is_timed_out(size) { - Err(TimeoutError::TimedOut(self.get_timeout(size))) - } else { - Ok(()) - } - } - - /// Get progress. - pub fn progress(&self) -> Option<&OperationProgress> { - self.progress.as_ref() - } - - /// Update progress. - pub fn update_progress(&self, bytes: u64) { - if let Some(ref progress) = self.progress { - progress.update(bytes); - } - } - - /// Check if operation is stalled (no progress for a while). - pub fn is_stalled(&self) -> bool { - self.progress.as_ref().is_some_and(|p| p.is_stale()) - } - - /// Get progress percentage. - pub fn progress_percent(&self) -> Option { - self.progress.as_ref().and_then(|p| p.progress_percent()) - } -} - -/// Timeout statistics. -#[derive(Debug, Default)] -pub struct TimeoutStats { - /// Total operations. - pub total_operations: AtomicU64, - /// Timed out operations. - pub timed_out: AtomicU64, - /// Total wait time in nanoseconds. - pub total_wait_time_ns: AtomicU64, - /// Maximum wait time in nanoseconds. - pub max_wait_time_ns: AtomicU64, -} - -impl TimeoutStats { - /// Create new timeout statistics. - pub fn new() -> Self { - Self::default() - } - - /// Record an operation. - pub fn record_operation(&self, wait_time: Duration) { - self.total_operations.fetch_add(1, Ordering::Relaxed); - let ns = wait_time.as_nanos() as u64; - self.total_wait_time_ns.fetch_add(ns, Ordering::Relaxed); - - let mut current = self.max_wait_time_ns.load(Ordering::Relaxed); - while ns > current { - match self - .max_wait_time_ns - .compare_exchange_weak(current, ns, Ordering::Relaxed, Ordering::Relaxed) - { - Ok(_) => break, - Err(actual) => current = actual, - } - } - } - - /// Record a timeout. - pub fn record_timeout(&self) { - self.timed_out.fetch_add(1, Ordering::Relaxed); - } - - /// Get timeout rate. - pub fn timeout_rate(&self) -> f64 { - let total = self.total_operations.load(Ordering::Relaxed); - let timed_out = self.timed_out.load(Ordering::Relaxed); - if total == 0 { 0.0 } else { timed_out as f64 / total as f64 } - } - - /// Get average wait time. - pub fn avg_wait_time(&self) -> Duration { - let total = self.total_wait_time_ns.load(Ordering::Relaxed); - let count = self.total_operations.load(Ordering::Relaxed); - total.checked_div(count).map(Duration::from_nanos).unwrap_or(Duration::ZERO) - } - - /// Reset statistics. - pub fn reset(&self) { - self.total_operations.store(0, Ordering::Relaxed); - self.timed_out.store(0, Ordering::Relaxed); - self.total_wait_time_ns.store(0, Ordering::Relaxed); - self.max_wait_time_ns.store(0, Ordering::Relaxed); - } -} - -/// Calculate adaptive timeout based on historical data and current conditions. -/// -/// This function adjusts the timeout based on: -/// - Historical transfer rate -/// - Recent timeout count -/// - Object size -pub fn calculate_adaptive_timeout( - base_timeout: Duration, - historical_rate_bps: Option, - recent_timeout_count: u32, - object_size: u64, -) -> Duration { - // If we have recent timeouts, increase timeout - let timeout_multiplier = if recent_timeout_count > 3 { - 2.0 // Double timeout if many recent timeouts - } else if recent_timeout_count > 1 { - 1.5 // 50% increase if some timeouts - } else { - 1.0 // No adjustment - }; - - // Adaptive timeout bounds: 5 seconds minimum, 10 minutes maximum. - const MIN_SECS: f64 = 5.0; - const MAX_SECS: f64 = 600.0; - - // If we have historical rate data, use it for estimation - let estimated_secs = match historical_rate_bps { - Some(rate) if rate > 0 => (object_size as f64 / rate as f64) * 1.2, // 20% buffer - _ => base_timeout.as_secs_f64(), - }; - - // Clamp BEFORE constructing the Duration: `from_secs_f64` panics when the - // estimate overflows Duration (huge object_size with a tiny historical rate). - Duration::from_secs_f64((estimated_secs * timeout_multiplier).clamp(MIN_SECS, MAX_SECS)) -} - -/// Estimate bytes per second transfer rate. -/// -/// This is used for adaptive timeout calculation. -pub fn estimate_bytes_per_second(object_size: u64, expected_duration: Duration) -> u64 { - let secs = expected_duration.as_secs_f64(); - if secs > 0.0 { - (object_size as f64 / secs) as u64 - } else { - // Return a reasonable default (1 MB/s) - 1024 * 1024 - } -} - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn test_timeout_config() { - let config = TimeoutConfig::default(); - assert!(config.validate().is_ok()); - - // Small file - let timeout = config.calculate_timeout(1024); - assert!(timeout >= config.min_timeout); - - // Large file - let timeout = config.calculate_timeout(100 * 1024 * 1024); - assert!(timeout <= config.max_timeout); - } - - #[test] - fn test_timeout_config_validation() { - let config = TimeoutConfig { - min_timeout: Duration::from_secs(10), - max_timeout: Duration::from_secs(5), - ..Default::default() - }; - assert!(config.validate().is_err()); - } - - #[test] - fn test_adaptive_timeout_extreme_estimate_does_not_panic() { - // A huge object with a tiny historical rate used to overflow - // Duration::from_secs_f64 and panic; it must clamp to the upper bound. - let timeout = calculate_adaptive_timeout(Duration::from_secs(30), Some(1), 0, u64::MAX); - assert_eq!(timeout, Duration::from_secs(600)); - - // Tiny estimates clamp to the lower bound. - let timeout = calculate_adaptive_timeout(Duration::from_secs(30), Some(u64::MAX), 0, 1); - assert_eq!(timeout, Duration::from_secs(5)); - } - - #[test] - fn test_operation_progress() { - let progress = OperationProgress::new(Some(1000), Duration::from_secs(5)); - - assert_eq!(progress.current(), 0); - assert_eq!(progress.progress_percent(), Some(0.0)); - - progress.update(500); - assert_eq!(progress.current(), 500); - assert_eq!(progress.progress_percent(), Some(50.0)); - - progress.add(300); - assert_eq!(progress.current(), 800); - assert_eq!(progress.remaining(), Some(200)); - } - - #[test] - fn test_request_timeout_wrapper() { - let config = TimeoutConfig { - base_timeout: Duration::from_millis(100), - enable_dynamic_timeout: false, - ..Default::default() - }; - let wrapper = RequestTimeoutWrapper::new(config); - - assert!(!wrapper.is_timed_out(None)); - - std::thread::sleep(Duration::from_millis(150)); - - assert!(wrapper.is_timed_out(None)); - assert!(wrapper.check_timeout(None).is_err()); - } - - #[test] - fn test_timeout_stats() { - let stats = TimeoutStats::new(); - - stats.record_operation(Duration::from_millis(10)); - stats.record_operation(Duration::from_millis(20)); - stats.record_timeout(); - - assert_eq!(stats.total_operations.load(Ordering::Relaxed), 2); - assert_eq!(stats.timed_out.load(Ordering::Relaxed), 1); - assert!((stats.timeout_rate() - 0.5).abs() < 0.01); - } - - #[test] - fn test_progress_tracking() { - let config = TimeoutConfig::default(); - let wrapper = RequestTimeoutWrapper::with_progress(config, Some(1000), Duration::from_secs(1)); - - wrapper.update_progress(500); - assert_eq!(wrapper.progress_percent(), Some(50.0)); - assert!(!wrapper.is_stalled()); - } -} diff --git a/crates/io-core/src/writer.rs b/crates/io-core/src/writer.rs deleted file mode 100644 index 2d1f1a754..000000000 --- a/crates/io-core/src/writer.rs +++ /dev/null @@ -1,443 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -//! BytesMut-backed object writer for optimized write operations. -//! -//! It uses `BytesMut` for efficient buffering; writes into that buffer may -//! still copy input bytes. The historical `ZeroCopyObjectWriter` name remains -//! available as a deprecated compatibility alias. - -use bytes::{BufMut, Bytes, BytesMut}; -use std::pin::Pin; -use std::task::{Context, Poll}; -use tokio::io::AsyncWrite; - -/// BytesMut-backed object writer for optimized write operations. -/// -/// This writer minimizes memory allocations by: -/// - Using BytesMut for efficient buffer growth -/// - Accepting `Bytes` inputs for efficient buffer handling -/// - Optional integration with BytesPool for buffer reuse -/// -/// # Example -/// -/// ```ignore -/// use rustfs_io_core::BytesMutWriter; -/// use bytes::Bytes; -/// -/// #[tokio::main] -/// async fn main() -> Result<(), Box> { -/// let mut writer = BytesMutWriter::new(); -/// -/// // Write into the internal BytesMut buffer -/// let data = Bytes::from("hello world"); -/// writer.write_buffered(data).await?; -/// -/// // Get the result as Bytes (zero-copy conversion) -/// let result = writer.into_bytes(); -/// -/// Ok(()) -/// } -/// ``` -pub struct BytesMutWriter { - /// Internal buffer using BytesMut for efficient growth - buffer: BytesMut, - /// Total bytes written - bytes_written: usize, - /// Whether the writer has been finalized - finalized: bool, -} - -/// Historical name for the BytesMut-backed object writer. -#[deprecated(since = "1.0.0-beta.8", note = "use BytesMutWriter; writes append into a BytesMut buffer")] -pub type ZeroCopyObjectWriter = BytesMutWriter; - -impl BytesMutWriter { - /// Create a new bytes-backed object writer with default capacity (8KB). - /// - /// # Example - /// - /// ```ignore - /// let writer = BytesMutWriter::new(); - /// ``` - pub fn new() -> Self { - Self::with_capacity(8 * 1024) - } - - /// Create a new bytes-backed object writer with specified capacity. - /// - /// # Arguments - /// - /// * `capacity` - Initial buffer capacity in bytes - /// - /// # Example - /// - /// ```ignore - /// let writer = BytesMutWriter::with_capacity(64 * 1024); - /// ``` - pub fn with_capacity(capacity: usize) -> Self { - Self { - buffer: BytesMut::with_capacity(capacity), - bytes_written: 0, - finalized: false, - } - } - - /// Write data into the internal buffer. - /// - /// This method accepts `Bytes` for API compatibility, then appends the - /// bytes into the internal `BytesMut` buffer. - /// - /// # Arguments - /// - /// * `data` - Data to append to the internal buffer - /// - /// # Returns - /// - /// * `Ok(usize)` - Number of bytes written - /// * `Err(ZeroCopyWriteError)` - Write error - /// - /// # Example - /// - /// ```ignore - /// let data = Bytes::from("hello world"); - /// let written = writer.write_buffered(data).await?; - /// ``` - pub async fn write_buffered(&mut self, data: Bytes) -> Result { - if self.finalized { - return Err(ZeroCopyWriteError::Finalized("Cannot write to finalized writer".to_string())); - } - - let len = data.len(); - self.buffer.put(data); - - self.bytes_written += len; - Ok(len) - } - - /// Historical name for `write_buffered`. - #[deprecated( - since = "1.0.0-beta.8", - note = "use write_buffered; this method appends bytes into an internal buffer" - )] - pub async fn write_zero_copy(&mut self, data: Bytes) -> Result { - self.write_buffered(data).await - } - - /// Write a slice of data. - /// - /// # Arguments - /// - /// * `data` - Data slice to write - /// - /// # Returns - /// - /// * `Ok(usize)` - Number of bytes written - /// * `Err(ZeroCopyWriteError)` - Write error - pub async fn write_slice(&mut self, data: &[u8]) -> Result { - if self.finalized { - return Err(ZeroCopyWriteError::Finalized("Cannot write to finalized writer".to_string())); - } - - let len = data.len(); - self.buffer.put_slice(data); - self.bytes_written += len; - Ok(len) - } - - /// Finalize the writer and consume it, returning the written data as Bytes. - /// - /// This converts the internal BytesMut to Bytes, which is a zero-copy - /// operation that freezes the buffer. - /// - /// # Returns - /// - /// The written data as Bytes - /// - /// # Example - /// - /// ```ignore - /// let result = writer.into_bytes(); - /// ``` - pub fn into_bytes(mut self) -> Bytes { - self.finalized = true; - self.buffer.freeze() - } - - /// Get the current buffer as a slice (without consuming). - /// - /// # Returns - /// - /// Slice of the current buffer content - pub fn as_slice(&self) -> &[u8] { - &self.buffer[..] - } - - /// Get the total number of bytes written. - /// - /// # Returns - /// - /// Number of bytes written - pub fn bytes_written(&self) -> usize { - self.bytes_written - } - - /// Get the current buffer capacity. - /// - /// # Returns - /// - /// Current buffer capacity in bytes - pub fn capacity(&self) -> usize { - self.buffer.capacity() - } - - /// Get the current buffer length. - /// - /// # Returns - /// - /// Current buffer length in bytes - pub fn len(&self) -> usize { - self.buffer.len() - } - - /// Check if the buffer is empty. - /// - /// # Returns - /// - /// `true` if buffer is empty, `false` otherwise - pub fn is_empty(&self) -> bool { - self.buffer.is_empty() - } - - /// Clear the buffer, resetting it to empty. - /// - /// This does not change the capacity, just resets the length to 0. - pub fn clear(&mut self) { - self.buffer.clear(); - self.bytes_written = 0; - self.finalized = false; - } - - /// Reserve additional capacity in the buffer. - /// - /// # Arguments - /// - /// * `additional` - Additional capacity to reserve - pub fn reserve(&mut self, additional: usize) { - self.buffer.reserve(additional); - } -} - -impl Default for BytesMutWriter { - fn default() -> Self { - Self::new() - } -} - -impl std::fmt::Debug for BytesMutWriter { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.debug_struct("BytesMutWriter") - .field("buffer_len", &self.buffer.len()) - .field("buffer_capacity", &self.buffer.capacity()) - .field("bytes_written", &self.bytes_written) - .field("finalized", &self.finalized) - .finish() - } -} - -/// AsyncWrite implementation for BytesMutWriter. -/// -/// This allows the writer to be used with tokio's async I/O utilities. -impl AsyncWrite for BytesMutWriter { - fn poll_write(mut self: Pin<&mut Self>, _cx: &mut Context<'_>, buf: &[u8]) -> Poll> { - if self.finalized { - return Poll::Ready(Err(tokio::io::Error::new( - tokio::io::ErrorKind::WriteZero, - "Cannot write to finalized writer", - ))); - } - - let len = buf.len(); - self.buffer.put_slice(buf); - self.bytes_written += len; - Poll::Ready(Ok(len)) - } - - fn poll_flush(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll> { - // Nothing to flush for in-memory buffer - Poll::Ready(Ok(())) - } - - fn poll_shutdown(mut self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll> { - self.finalized = true; - Poll::Ready(Ok(())) - } -} - -/// Zero-copy write error types. -#[derive(Debug, thiserror::Error)] -pub enum ZeroCopyWriteError { - /// I/O error occurred - #[error("I/O error: {0}")] - Io(#[from] tokio::io::Error), - - /// Writer has been finalized and cannot accept more writes - #[error("Writer finalized: {0}")] - Finalized(String), - - /// Invalid input provided - #[error("Invalid input: {0}")] - InvalidInput(String), -} - -#[cfg(test)] -mod tests { - use super::*; - - #[tokio::test] - async fn test_new_writer() { - let writer = BytesMutWriter::new(); - assert!(writer.is_empty()); - assert_eq!(writer.bytes_written(), 0); - assert!(writer.capacity() >= 8 * 1024); - } - - #[tokio::test] - async fn test_write_buffered() { - let mut writer = BytesMutWriter::new(); - let data = Bytes::from("hello world"); - - let written = writer.write_buffered(data).await.unwrap(); - assert_eq!(written, 11); - assert_eq!(writer.bytes_written(), 11); - assert_eq!(writer.as_slice(), b"hello world"); - } - - #[tokio::test] - async fn test_preferred_writer_alias() { - let mut writer = BytesMutWriter::new(); - let written = writer - .write_buffered(Bytes::from("hello world")) - .await - .expect("write bytes through alias"); - - assert_eq!(written, 11); - assert_eq!(writer.as_slice(), b"hello world"); - } - - #[tokio::test] - async fn test_write_slice() { - let mut writer = BytesMutWriter::new(); - let data = b"hello world"; - - let written = writer.write_slice(data).await.unwrap(); - assert_eq!(written, 11); - assert_eq!(writer.bytes_written(), 11); - assert_eq!(writer.as_slice(), b"hello world"); - } - - #[tokio::test] - async fn test_into_bytes() { - let mut writer = BytesMutWriter::new(); - let data = Bytes::from("hello world"); - - writer.write_buffered(data).await.unwrap(); - let result = writer.into_bytes(); - - assert_eq!(result.as_ref(), b"hello world"); - } - - #[tokio::test] - async fn test_write_after_finalize() { - let mut writer = BytesMutWriter::new(); - let data = Bytes::from("hello"); - - writer.write_buffered(data).await.unwrap(); - let _result = writer.into_bytes(); - - // Create new writer and try to write after finalize - let mut writer2 = BytesMutWriter::new(); - writer2.write_buffered(Bytes::from("test")).await.unwrap(); - let _ = writer2.into_bytes(); - - // Writing to a consumed writer should work via new writer - let mut writer3 = BytesMutWriter::new(); - let result = writer3.write_buffered(Bytes::from("final")).await; - assert!(result.is_ok()); - } - - #[tokio::test] - async fn test_clear() { - let mut writer = BytesMutWriter::new(); - writer.write_slice(b"hello").await.unwrap(); - - writer.clear(); - assert!(writer.is_empty()); - assert_eq!(writer.bytes_written(), 0); - // Capacity should remain - assert!(writer.capacity() > 0); - } - - #[tokio::test] - async fn test_reserve() { - let mut writer = BytesMutWriter::with_capacity(10); - let initial_capacity = writer.capacity(); - - writer.reserve(1000); - // Reserve ensures at least the additional capacity can be added - // but may allocate more than requested - assert!(writer.capacity() >= initial_capacity); - } - - #[tokio::test] - async fn test_multiple_writes() { - let mut writer = BytesMutWriter::new(); - - writer.write_buffered(Bytes::from("hello ")).await.unwrap(); - writer.write_slice(b"world").await.unwrap(); - - assert_eq!(writer.as_slice(), b"hello world"); - assert_eq!(writer.bytes_written(), 11); - } - - #[tokio::test] - async fn test_async_write() { - use tokio::io::AsyncWriteExt; - - let mut writer = BytesMutWriter::new(); - let data = b"hello world"; - - let written = writer.write(data).await.unwrap(); - assert_eq!(written, 11); - assert_eq!(writer.as_slice(), b"hello world"); - } - - #[tokio::test] - async fn test_debug() { - let writer = BytesMutWriter::new(); - let debug_str = format!("{:?}", writer); - assert!(debug_str.contains("BytesMutWriter")); - assert!(debug_str.contains("buffer_len")); - } - - #[tokio::test] - #[allow(deprecated)] - async fn test_legacy_writer_alias() { - let mut writer = ZeroCopyObjectWriter::new(); - let written = writer.write_zero_copy(Bytes::from("hello")).await.unwrap(); - - assert_eq!(written, 5); - assert_eq!(writer.as_slice(), b"hello"); - } -} From f06a9c9cbacb826ef4ce58a1e8b2da954e6c1044 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 17:04:55 +0800 Subject: [PATCH 11/24] fix(deps): record heal's bytes and crc-fast in the lockfile (#6211) From 68547ed7ea4a325e4b148778dfbdf188269700e0 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 17:05:30 +0800 Subject: [PATCH 12/24] test(e2e): drop 187 no-op serial markers from the three densest e2e suites (#6209) test(e2e): drop no-op serial markers from the three densest e2e suites serial_test's #[serial] is an in-process mutex. cargo-nextest, this repo's authoritative runner, executes every test in its own process, so the mutex is never contended and the attribute is a documented no-op -- see the "Serial execution & nextest profiles" section of docs/testing/README.md and the header of .config/nextest.toml. Cross-process serialization is provided only by a [test-groups] entry with max-threads = 1. Remove 187 such markers (plus 3 now-unused imports) from the three marker-densest modules of crates/e2e_test: multipart_auth_test.rs 85 replication_extension_test.rs 68 object_lock/object_lock_test.rs 34 None of these modules is covered by any [test-groups] entry, so the markers were carrying no isolation for any lane. Every test in all three files builds its own server via RustFSTestEnvironment::new(), which gives a UUID temp dir and a uniquely allocated port -- the .config/nextest.toml comment on replication_extension_test already states this explicitly ("parallel-safe by construction"). No test mutates process env, binds a fixed port, or touches process-global state, so nothing here needed temp_env or a test-group instead. Pure deletion: 190 lines removed, 0 added, no test renamed, no behaviour changed. serial_test stays in Cargo.toml -- 338 markers across 90 other files in the crate still use it. Refs: backlog#1846 (T1). --- crates/e2e_test/src/multipart_auth_test.rs | 86 ------------------- .../src/object_lock/object_lock_test.rs | 35 -------- .../src/replication_extension_test.rs | 69 --------------- 3 files changed, 190 deletions(-) diff --git a/crates/e2e_test/src/multipart_auth_test.rs b/crates/e2e_test/src/multipart_auth_test.rs index 81613bf97..dc0d6dd33 100644 --- a/crates/e2e_test/src/multipart_auth_test.rs +++ b/crates/e2e_test/src/multipart_auth_test.rs @@ -30,7 +30,6 @@ use md5::{Digest as Md5Digest, Md5}; use rustfs_signer::constants::UNSIGNED_PAYLOAD; use rustfs_signer::sign_v4; use s3s::Body; -use serial_test::serial; use std::collections::HashMap; use std::error::Error; use std::io::Cursor; @@ -356,7 +355,6 @@ async fn run_post_object_policy_case( /// smuggles one extra field the policy never declared, and the upload must be /// rejected with 403 AccessDenied naming the offending field. #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_fields_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -484,7 +482,6 @@ async fn test_anonymous_post_object_rejects_fields_missing_from_policy_condition /// sends a different one, and the upload must be rejected with 400 /// InvalidPolicyDocument naming the field. #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_exact_condition_policy_mismatches() -> Result<(), Box> { init_logging(); @@ -689,7 +686,6 @@ async fn test_anonymous_post_object_rejects_exact_condition_policy_mismatches() /// one of them with a different value, and the upload must be rejected with /// 400 InvalidPolicyDocument naming the mismatched field. #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_policy_mismatches() -> Result<(), Box> { init_logging(); @@ -757,7 +753,6 @@ async fn test_anonymous_post_object_rejects_object_lock_policy_mismatches() -> R /// exact values, the form sends a different parameter value, and the upload /// must be rejected with 400 InvalidPolicyDocument naming the parameter. #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_sse_kms_policy_mismatches() -> Result<(), Box> { init_logging(); @@ -839,7 +834,6 @@ async fn test_anonymous_post_object_rejects_sse_kms_policy_mismatches() -> Resul /// NotImplemented (SSE-KMS POST uploads are not implemented), not with a /// policy error. #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_sse_kms_params_outside_policy_conditions() -> Result<(), Box> { init_logging(); @@ -894,7 +888,6 @@ async fn test_anonymous_post_object_rejects_sse_kms_params_outside_policy_condit } #[tokio::test] -#[serial] async fn test_anonymous_multipart_control_apis_require_auth() -> Result<(), Box> { init_logging(); @@ -968,7 +961,6 @@ async fn test_anonymous_multipart_control_apis_require_auth() -> Result<(), Box< } #[tokio::test] -#[serial] async fn test_anonymous_post_object_requires_auth() -> Result<(), Box> { init_logging(); @@ -1002,7 +994,6 @@ async fn test_anonymous_post_object_requires_auth() -> Result<(), Box Result<(), Box> { init_logging(); @@ -1066,7 +1057,6 @@ async fn test_anonymous_post_object_honors_success_action_status() -> Result<(), } #[tokio::test] -#[serial] async fn test_anonymous_post_object_honors_success_action_redirect() -> Result<(), Box> { init_logging(); @@ -1139,7 +1129,6 @@ async fn test_anonymous_post_object_honors_success_action_redirect() -> Result<( } #[tokio::test] -#[serial] async fn test_anonymous_post_object_defaults_to_no_content() -> Result<(), Box> { init_logging(); @@ -1185,7 +1174,6 @@ async fn test_anonymous_post_object_defaults_to_no_content() -> Result<(), Box Result<(), Box> { init_logging(); @@ -1232,7 +1220,6 @@ async fn test_anonymous_post_object_rejects_sse_kms() -> Result<(), Box Result<(), Box> { init_logging(); @@ -1290,7 +1277,6 @@ async fn test_anonymous_post_object_accepts_sse_s3() -> Result<(), Box Result<(), Box> { init_logging(); @@ -1363,7 +1349,6 @@ async fn test_anonymous_post_object_uses_bucket_default_sse_s3() -> Result<(), B } #[tokio::test] -#[serial] async fn test_anonymous_post_object_uses_bucket_default_sse_kms() -> Result<(), Box> { init_logging(); @@ -1437,7 +1422,6 @@ async fn test_anonymous_post_object_uses_bucket_default_sse_kms() -> Result<(), } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_sse_s3_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -1488,7 +1472,6 @@ async fn test_anonymous_post_object_rejects_sse_s3_policy_mismatch() -> Result<( } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_sse_s3_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -1552,7 +1535,6 @@ async fn test_anonymous_post_object_accepts_sse_s3_missing_from_policy_condition } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_storage_class_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -1606,7 +1588,6 @@ async fn test_anonymous_post_object_accepts_storage_class_exact_policy_match() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_storage_class_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -1657,7 +1638,6 @@ async fn test_anonymous_post_object_rejects_storage_class_missing_from_policy_co } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_invalid_storage_class_value() -> Result<(), Box> { init_logging(); @@ -1709,7 +1689,6 @@ async fn test_anonymous_post_object_rejects_invalid_storage_class_value() -> Res } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_checksum_algorithm_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -1765,7 +1744,6 @@ async fn test_anonymous_post_object_rejects_checksum_algorithm_missing_from_poli } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_checksum_algorithm_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -1822,7 +1800,6 @@ async fn test_anonymous_post_object_rejects_checksum_algorithm_policy_mismatch() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_checksum_auxiliary_fields_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -1886,7 +1863,6 @@ async fn test_anonymous_post_object_rejects_checksum_auxiliary_fields_missing_fr } #[tokio::test] -#[serial] async fn test_anonymous_post_object_allows_sse_c_fields_outside_policy_conditions() -> Result<(), Box> { init_logging(); @@ -1963,7 +1939,6 @@ async fn test_anonymous_post_object_allows_sse_c_fields_outside_policy_condition } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_sse_c_exact_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -2022,7 +1997,6 @@ async fn test_anonymous_post_object_rejects_sse_c_exact_policy_mismatch() -> Res } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_duplicate_key_form_values() -> Result<(), Box> { init_logging(); @@ -2072,7 +2046,6 @@ async fn test_anonymous_post_object_rejects_duplicate_key_form_values() -> Resul } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_invalid_success_action_status() -> Result<(), Box> { init_logging(); @@ -2120,7 +2093,6 @@ async fn test_anonymous_post_object_rejects_invalid_success_action_status() -> R } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_invalid_success_action_redirect() -> Result<(), Box> { init_logging(); @@ -2168,7 +2140,6 @@ async fn test_anonymous_post_object_rejects_invalid_success_action_redirect() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_form_fields_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -2223,7 +2194,6 @@ async fn test_anonymous_post_object_rejects_form_fields_missing_from_policy_cond } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_form_fields_covered_by_policy_conditions() -> Result<(), Box> { init_logging(); @@ -2280,7 +2250,6 @@ async fn test_anonymous_post_object_accepts_form_fields_covered_by_policy_condit } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_starts_with_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -2335,7 +2304,6 @@ async fn test_anonymous_post_object_rejects_starts_with_policy_mismatch() -> Res } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_content_length_range_violation() -> Result<(), Box> { init_logging(); @@ -2388,7 +2356,6 @@ async fn test_anonymous_post_object_rejects_content_length_range_violation() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_success_action_status_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2445,7 +2412,6 @@ async fn test_anonymous_post_object_accepts_success_action_status_exact_policy_m } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_success_action_redirect_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -2502,7 +2468,6 @@ async fn test_anonymous_post_object_rejects_success_action_redirect_policy_misma } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_success_action_redirect_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2568,7 +2533,6 @@ async fn test_anonymous_post_object_accepts_success_action_redirect_exact_policy } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_success_action_redirect_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -2621,7 +2585,6 @@ async fn test_anonymous_post_object_rejects_success_action_redirect_missing_from } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_metadata_field_covered_by_starts_with() -> Result<(), Box> { init_logging(); @@ -2676,7 +2639,6 @@ async fn test_anonymous_post_object_accepts_metadata_field_covered_by_starts_wit } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_content_type_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2734,7 +2696,6 @@ async fn test_anonymous_post_object_accepts_content_type_field_exact_policy_matc } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_content_type_field_covered_by_starts_with() -> Result<(), Box> { init_logging(); @@ -2792,7 +2753,6 @@ async fn test_anonymous_post_object_accepts_content_type_field_covered_by_starts } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_content_disposition_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2850,7 +2810,6 @@ async fn test_anonymous_post_object_accepts_content_disposition_field_exact_poli } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_cache_control_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2908,7 +2867,6 @@ async fn test_anonymous_post_object_accepts_cache_control_field_exact_policy_mat } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_content_language_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -2966,7 +2924,6 @@ async fn test_anonymous_post_object_accepts_content_language_field_exact_policy_ } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_content_encoding_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -3024,7 +2981,6 @@ async fn test_anonymous_post_object_accepts_content_encoding_field_exact_policy_ } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_website_redirect_location_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -3082,7 +3038,6 @@ async fn test_anonymous_post_object_accepts_website_redirect_location_exact_poli } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_expires_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -3140,7 +3095,6 @@ async fn test_anonymous_post_object_accepts_expires_field_exact_policy_match() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_retention_without_permission() -> Result<(), Box> { init_logging(); @@ -3196,7 +3150,6 @@ async fn test_anonymous_post_object_rejects_object_lock_retention_without_permis } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_retention_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -3256,7 +3209,6 @@ async fn test_anonymous_post_object_rejects_object_lock_retention_missing_from_p } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_legal_hold_without_permission() -> Result<(), Box> { init_logging(); @@ -3309,7 +3261,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_without_permi } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_legal_hold_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -3368,7 +3319,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_policy_mismat } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_object_lock_legal_hold_missing_from_policy_conditions() -> Result<(), Box> { init_logging(); @@ -3426,7 +3376,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_missing_from_ } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_tagging_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -3492,7 +3441,6 @@ async fn test_anonymous_post_object_accepts_tagging_field_exact_policy_match() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_accepts_metadata_field_exact_policy_match() -> Result<(), Box> { init_logging(); @@ -3551,7 +3499,6 @@ async fn test_anonymous_post_object_accepts_metadata_field_exact_policy_match() } #[tokio::test] -#[serial] async fn test_anonymous_post_object_allows_x_ignore_fields_outside_policy_conditions() -> Result<(), Box> { init_logging(); @@ -3604,7 +3551,6 @@ async fn test_anonymous_post_object_allows_x_ignore_fields_outside_policy_condit } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_sigv4_date_policy_mismatch() -> Result<(), Box> { init_logging(); @@ -3657,7 +3603,6 @@ async fn test_anonymous_post_object_rejects_sigv4_date_policy_mismatch() -> Resu } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_mismatched_bucket_form_field() -> Result<(), Box> { init_logging(); @@ -3712,7 +3657,6 @@ async fn test_anonymous_post_object_rejects_mismatched_bucket_form_field() -> Re } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_multiple_bucket_values() -> Result<(), Box> { init_logging(); @@ -3764,7 +3708,6 @@ async fn test_anonymous_post_object_rejects_multiple_bucket_values() -> Result<( } #[tokio::test] -#[serial] async fn test_anonymous_post_object_rejects_extra_content_disposition_field() -> Result<(), Box> { init_logging(); @@ -3820,7 +3763,6 @@ async fn test_anonymous_post_object_rejects_extra_content_disposition_field() } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_tar_entries_with_prefix_headers() -> Result<(), Box> { init_logging(); @@ -3891,7 +3833,6 @@ async fn test_signed_put_object_extract_expands_tar_entries_with_prefix_headers( } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_objects() -> Result<(), Box> { init_logging(); @@ -3956,7 +3897,6 @@ async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_ } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_sse_s3_and_redirect() -> Result<(), Box> { init_logging(); @@ -4004,7 +3944,6 @@ async fn test_signed_put_object_extract_preserves_sse_s3_and_redirect() -> Resul } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_storage_class() -> Result<(), Box> { init_logging(); @@ -4047,7 +3986,6 @@ async fn test_signed_put_object_extract_preserves_storage_class() -> Result<(), } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_rejects_invalid_storage_class() -> Result<(), Box> { init_logging(); @@ -4083,7 +4021,6 @@ async fn test_signed_put_object_extract_rejects_invalid_storage_class() -> Resul } #[tokio::test] -#[serial] async fn test_signed_put_object_rejects_write_offset_bytes_header() -> Result<(), Box> { init_logging(); @@ -4137,7 +4074,6 @@ async fn test_signed_put_object_rejects_write_offset_bytes_header() -> Result<() } #[tokio::test] -#[serial] async fn test_raw_signed_put_object_write_offset_bytes_returns_minio_compatible_error_body() -> Result<(), Box> { init_logging(); @@ -4176,7 +4112,6 @@ async fn test_raw_signed_put_object_write_offset_bytes_returns_minio_compatible_ } #[tokio::test] -#[serial] async fn test_anonymous_put_object_write_offset_bytes_returns_minio_compatible_error_body() -> Result<(), Box> { init_logging(); @@ -4235,7 +4170,6 @@ async fn test_anonymous_put_object_write_offset_bytes_returns_minio_compatible_e } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_uses_bucket_default_sse_s3() -> Result<(), Box> { init_logging(); @@ -4300,7 +4234,6 @@ async fn test_signed_put_object_extract_uses_bucket_default_sse_s3() -> Result<( } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_rejects_bucket_default_sse_kms() -> Result<(), Box> { init_logging(); @@ -4356,7 +4289,6 @@ async fn test_signed_put_object_extract_rejects_bucket_default_sse_kms() -> Resu } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_sse_c() -> Result<(), Box> { init_logging(); @@ -4421,7 +4353,6 @@ async fn test_signed_put_object_extract_preserves_sse_c() -> Result<(), Box Result<(), Box> { init_logging(); @@ -4476,7 +4407,6 @@ async fn test_signed_put_object_extract_preserves_object_lock_legal_hold() -> Re } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_object_lock_retention() -> Result<(), Box> { init_logging(); @@ -4536,7 +4466,6 @@ async fn test_signed_put_object_extract_preserves_object_lock_retention() -> Res } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_pax_retention_overrides_request_retention() -> Result<(), Box> { init_logging(); @@ -4600,7 +4529,6 @@ async fn test_signed_put_object_extract_pax_retention_overrides_request_retentio } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_returns_archive_etag() -> Result<(), Box> { init_logging(); @@ -4634,7 +4562,6 @@ async fn test_signed_put_object_extract_returns_archive_etag() -> Result<(), Box } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_entry_mtime() -> Result<(), Box> { init_logging(); @@ -4670,7 +4597,6 @@ async fn test_signed_put_object_extract_preserves_entry_mtime() -> Result<(), Bo } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_pax_metadata_and_version_id() -> Result<(), Box> { init_logging(); @@ -4724,7 +4650,6 @@ async fn test_signed_put_object_extract_preserves_pax_metadata_and_version_id() } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_authorizes_each_pax_privilege_and_retention_conditions() -> Result<(), Box> { init_logging(); @@ -5034,7 +4959,6 @@ async fn test_signed_put_object_extract_authorizes_each_pax_privilege_and_retent } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_accepts_compat_header() -> Result<(), Box> { init_logging(); @@ -5076,7 +5000,6 @@ async fn test_signed_put_object_extract_accepts_compat_header() -> Result<(), Bo } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_preserves_directory_markers_by_default() -> Result<(), Box> { init_logging(); @@ -5137,7 +5060,6 @@ async fn test_signed_put_object_extract_preserves_directory_markers_by_default() } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_tar_gz_archive() -> Result<(), Box> { init_logging(); @@ -5189,7 +5111,6 @@ async fn test_signed_put_object_extract_expands_tar_gz_archive() -> Result<(), B } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_tgz_archive() -> Result<(), Box> { init_logging(); @@ -5241,7 +5162,6 @@ async fn test_signed_put_object_extract_expands_tgz_archive() -> Result<(), Box< } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_tbz2_archive() -> Result<(), Box> { init_logging(); @@ -5293,7 +5213,6 @@ async fn test_signed_put_object_extract_expands_tbz2_archive() -> Result<(), Box } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_txz_archive() -> Result<(), Box> { init_logging(); @@ -5345,7 +5264,6 @@ async fn test_signed_put_object_extract_expands_txz_archive() -> Result<(), Box< } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_enabled() -> Result<(), Box> { init_logging(); @@ -5419,7 +5337,6 @@ async fn test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_e } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_normalizes_prefix_header_value() -> Result<(), Box> { init_logging(); @@ -5462,7 +5379,6 @@ async fn test_signed_put_object_extract_normalizes_prefix_header_value() -> Resu } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_expands_tzst_archive() -> Result<(), Box> { init_logging(); @@ -5514,7 +5430,6 @@ async fn test_signed_put_object_extract_expands_tzst_archive() -> Result<(), Box } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_rejects_missing_archive_extension() -> Result<(), Box> { init_logging(); @@ -5548,7 +5463,6 @@ async fn test_signed_put_object_extract_rejects_missing_archive_extension() -> R } #[tokio::test] -#[serial] async fn test_signed_put_object_extract_rejects_invalid_tar_gz_payload() -> Result<(), Box> { init_logging(); diff --git a/crates/e2e_test/src/object_lock/object_lock_test.rs b/crates/e2e_test/src/object_lock/object_lock_test.rs index 0748ad499..7a9d6f9bc 100644 --- a/crates/e2e_test/src/object_lock/object_lock_test.rs +++ b/crates/e2e_test/src/object_lock/object_lock_test.rs @@ -33,7 +33,6 @@ use aws_sdk_s3::types::{ ObjectLockMode, ObjectLockRetentionMode, }; use chrono::{DateTime, Duration, Utc}; -use serial_test::serial; use tracing::info; /// Initialize test logging @@ -107,7 +106,6 @@ fn parse_s3_datetime(value: &aws_sdk_s3::primitives::DateTime) -> DateTime // ============================================================================ #[tokio::test] -#[serial] async fn test_delete_object_blocked_by_compliance_retention() { init_logging(); info!("🧪 Test: DeleteObject blocked by COMPLIANCE retention"); @@ -145,7 +143,6 @@ async fn test_delete_object_blocked_by_compliance_retention() { } #[tokio::test] -#[serial] async fn test_delete_object_blocked_by_governance_without_bypass() { init_logging(); info!("🧪 Test: DeleteObject blocked by GOVERNANCE retention without bypass"); @@ -175,7 +172,6 @@ async fn test_delete_object_blocked_by_governance_without_bypass() { } #[tokio::test] -#[serial] async fn test_delete_object_allowed_by_governance_with_bypass() { init_logging(); info!("🧪 Test: DeleteObject allowed by GOVERNANCE retention with bypass"); @@ -215,7 +211,6 @@ async fn test_delete_object_allowed_by_governance_with_bypass() { } #[tokio::test] -#[serial] async fn test_delete_object_creates_delete_marker_for_retained_current_version() { init_logging(); info!("🧪 Test: DeleteObject creates delete marker for retained current version"); @@ -266,7 +261,6 @@ async fn test_delete_object_creates_delete_marker_for_retained_current_version() } #[tokio::test] -#[serial] async fn test_delete_object_blocked_by_legal_hold() { init_logging(); info!("🧪 Test: DeleteObject blocked by Legal Hold"); @@ -299,7 +293,6 @@ async fn test_delete_object_blocked_by_legal_hold() { } #[tokio::test] -#[serial] async fn test_delete_object_allowed_with_legal_hold_off() { init_logging(); info!("🧪 Test: DeleteObject allowed with Legal Hold OFF"); @@ -335,7 +328,6 @@ async fn test_delete_object_allowed_with_legal_hold_off() { } #[tokio::test] -#[serial] async fn test_delete_object_after_legal_hold_removed() { init_logging(); info!("🧪 Test: DeleteObject succeeds after Legal Hold is removed"); @@ -369,7 +361,6 @@ async fn test_delete_object_after_legal_hold_removed() { } #[tokio::test] -#[serial] async fn test_get_object_legal_hold_returns_updated_status() { init_logging(); info!("🧪 Test: GetObjectLegalHold returns updated status"); @@ -425,7 +416,6 @@ async fn test_get_object_legal_hold_returns_updated_status() { } #[tokio::test] -#[serial] async fn test_get_object_retention_returns_configured_values() { init_logging(); info!("🧪 Test: GetObjectRetention returns configured values"); @@ -476,7 +466,6 @@ async fn test_get_object_retention_returns_configured_values() { // creating a new current version. The lock protects the existing version // from deletion; it never blocks new versions. #[tokio::test] -#[serial] async fn test_put_object_overwrite_creates_new_version_under_legal_hold() { init_logging(); info!("🧪 Test: PutObject overwrite of a legal-hold version creates a new version"); @@ -561,7 +550,6 @@ async fn test_put_object_overwrite_creates_new_version_under_legal_hold() { } #[tokio::test] -#[serial] async fn test_copy_object_applies_requested_legal_hold() { init_logging(); info!("🧪 Test: CopyObject applies requested Legal Hold"); @@ -613,7 +601,6 @@ async fn test_copy_object_applies_requested_legal_hold() { } #[tokio::test] -#[serial] async fn test_copy_object_does_not_inherit_source_legal_hold() { init_logging(); info!("🧪 Test: CopyObject does not inherit source Legal Hold"); @@ -707,7 +694,6 @@ async fn test_copy_object_does_not_inherit_source_legal_hold() { } #[tokio::test] -#[serial] async fn test_copy_object_overwrite_creates_new_version_under_legal_hold() { init_logging(); info!("🧪 Test: CopyObject overwrite of a legal-hold destination creates a new version"); @@ -787,7 +773,6 @@ async fn test_copy_object_overwrite_creates_new_version_under_legal_hold() { } #[tokio::test] -#[serial] async fn test_create_multipart_upload_applies_requested_legal_hold() { init_logging(); info!("🧪 Test: CreateMultipartUpload applies requested Legal Hold"); @@ -853,7 +838,6 @@ async fn test_create_multipart_upload_applies_requested_legal_hold() { } #[tokio::test] -#[serial] async fn test_create_multipart_upload_creates_new_version_under_compliance_retention() { init_logging(); info!("🧪 Test: CreateMultipartUpload over a COMPLIANCE-retained key creates a new version"); @@ -933,7 +917,6 @@ async fn test_create_multipart_upload_creates_new_version_under_compliance_reten } #[tokio::test] -#[serial] async fn test_delete_completed_multipart_object_blocked_by_legal_hold() { init_logging(); info!("🧪 Test: Delete completed multipart object blocked by Legal Hold"); @@ -993,7 +976,6 @@ async fn test_delete_completed_multipart_object_blocked_by_legal_hold() { } #[tokio::test] -#[serial] async fn test_delete_completed_multipart_object_blocked_by_retention() { init_logging(); info!("🧪 Test: Delete completed multipart object blocked by retention"); @@ -1055,7 +1037,6 @@ async fn test_delete_completed_multipart_object_blocked_by_retention() { } #[tokio::test] -#[serial] async fn test_complete_multipart_upload_creates_new_version_under_legal_hold() { init_logging(); info!("🧪 Test: CompleteMultipartUpload creates a new version when the current version is under Legal Hold"); @@ -1135,7 +1116,6 @@ async fn test_complete_multipart_upload_creates_new_version_under_legal_hold() { } #[tokio::test] -#[serial] async fn test_complete_multipart_upload_creates_new_version_under_compliance_retention() { init_logging(); info!("🧪 Test: CompleteMultipartUpload creates a new version when the current version is under COMPLIANCE retention"); @@ -1209,7 +1189,6 @@ async fn test_complete_multipart_upload_creates_new_version_under_compliance_ret } #[tokio::test] -#[serial] async fn test_write_paths_require_put_object_legal_hold_permission() { init_logging(); info!("🧪 Test: write paths require PutObjectLegalHold permission"); @@ -1273,7 +1252,6 @@ async fn test_write_paths_require_put_object_legal_hold_permission() { } #[tokio::test] -#[serial] async fn test_write_paths_require_put_object_retention_permission() { init_logging(); info!("🧪 Test: write paths require PutObjectRetention permission"); @@ -1345,7 +1323,6 @@ async fn test_write_paths_require_put_object_retention_permission() { // ============================================================================ #[tokio::test] -#[serial] async fn test_delete_objects_mixed_locked_unlocked() { init_logging(); info!("🧪 Test: DeleteObjects with mixed locked and unlocked objects"); @@ -1427,7 +1404,6 @@ async fn test_delete_objects_mixed_locked_unlocked() { // ============================================================================ #[tokio::test] -#[serial] async fn test_put_retention_compliance_cannot_shorten() { init_logging(); info!("🧪 Test: PutObjectRetention cannot shorten COMPLIANCE retention"); @@ -1468,7 +1444,6 @@ async fn test_put_retention_compliance_cannot_shorten() { } #[tokio::test] -#[serial] async fn test_put_retention_compliance_can_extend() { init_logging(); info!("🧪 Test: PutObjectRetention can extend COMPLIANCE retention"); @@ -1509,7 +1484,6 @@ async fn test_put_retention_compliance_can_extend() { } #[tokio::test] -#[serial] async fn test_put_retention_governance_extend_without_bypass() { init_logging(); info!("🧪 Test: PutObjectRetention on GOVERNANCE can extend without bypass"); @@ -1553,7 +1527,6 @@ async fn test_put_retention_governance_extend_without_bypass() { } #[tokio::test] -#[serial] async fn test_put_retention_governance_shorten_requires_bypass() { init_logging(); info!("🧪 Test: PutObjectRetention on GOVERNANCE requires bypass to shorten"); @@ -1615,7 +1588,6 @@ async fn test_put_retention_governance_shorten_requires_bypass() { // ============================================================================ #[tokio::test] -#[serial] async fn test_default_retention_applied_to_new_objects() { init_logging(); info!("🧪 Test: Default retention is applied to new objects"); @@ -1685,7 +1657,6 @@ async fn test_default_retention_applied_to_new_objects() { } #[tokio::test] -#[serial] async fn test_delete_object_creates_delete_marker_for_default_retained_current_version() { init_logging(); info!("🧪 Test: DeleteObject creates delete marker for default-retained current version"); @@ -1770,7 +1741,6 @@ async fn test_delete_object_creates_delete_marker_for_default_retained_current_v } #[tokio::test] -#[serial] async fn test_put_copy_and_multipart_reject_incomplete_retention_headers() { init_logging(); info!("🧪 Test: write paths reject incomplete Object Lock retention headers"); @@ -1869,7 +1839,6 @@ async fn test_put_copy_and_multipart_reject_incomplete_retention_headers() { } #[tokio::test] -#[serial] async fn test_copy_object_retention_uses_destination_policy() { init_logging(); info!("🧪 Test: CopyObject retention follows destination policy"); @@ -2051,7 +2020,6 @@ async fn test_copy_object_retention_uses_destination_policy() { } #[tokio::test] -#[serial] async fn test_multipart_default_retention_fixed_at_create() { init_logging(); info!("🧪 Test: multipart default retention is fixed at CreateMultipartUpload"); @@ -2122,7 +2090,6 @@ async fn test_multipart_default_retention_fixed_at_create() { // ============================================================================ #[tokio::test] -#[serial] async fn test_unretained_object_lock_object_delete_and_bucket_cleanup() { init_logging(); info!("🧪 Test: Unretained Object Lock object delete and bucket cleanup (Issue #5339)"); @@ -2243,7 +2210,6 @@ async fn test_unretained_object_lock_object_delete_and_bucket_cleanup() { } #[tokio::test] -#[serial] async fn test_versioning_auto_enabled_with_object_lock() { init_logging(); info!("🧪 Test: Versioning is auto-enabled when Object Lock is configured"); @@ -2302,7 +2268,6 @@ async fn test_versioning_auto_enabled_with_object_lock() { // ============================================================================ #[tokio::test] -#[serial] async fn test_error_message_distinguishes_legal_hold_from_retention() { init_logging(); info!("🧪 Test: Error messages distinguish Legal Hold from Retention"); diff --git a/crates/e2e_test/src/replication_extension_test.rs b/crates/e2e_test/src/replication_extension_test.rs index 1941bb105..ec847bfb7 100644 --- a/crates/e2e_test/src/replication_extension_test.rs +++ b/crates/e2e_test/src/replication_extension_test.rs @@ -60,7 +60,6 @@ use rustfs_signer::constants::UNSIGNED_PAYLOAD; use rustfs_signer::sign_v4; use s3s::Body; use s3s::header::X_AMZ_REPLICATION_STATUS; -use serial_test::serial; use sha2::{Digest, Sha256}; use std::collections::BTreeMap; use std::convert::Infallible; @@ -2506,7 +2505,6 @@ async fn build_replication_pair( /// metadata was inherited wholesale from the source, so the scanner heal pass /// skipped it too — no PENDING/FAILED marker meant nothing to re-drive). #[tokio::test] -#[serial] async fn test_copy_object_replicates_to_target() -> TestResult { init_logging(); @@ -2555,7 +2553,6 @@ async fn test_copy_object_replicates_to_target() -> TestResult { /// independent object; every member must replicate to the remote target like a /// regular PUT (MinIO PutObjectExtract parity). #[tokio::test] -#[serial] async fn test_snowball_extract_replicates_members_to_target() -> TestResult { init_logging(); @@ -2601,7 +2598,6 @@ async fn test_snowball_extract_replicates_members_to_target() -> TestResult { } #[tokio::test] -#[serial] async fn test_replication_check_succeeds_with_remote_target() -> Result<(), Box> { init_logging(); @@ -2638,7 +2634,6 @@ async fn test_replication_check_succeeds_with_remote_target() -> Result<(), Box< } #[tokio::test] -#[serial] async fn test_replication_check_rejects_target_without_object_lock() -> Result<(), Box> { init_logging(); @@ -2692,7 +2687,6 @@ async fn test_replication_check_rejects_target_without_object_lock() -> Result<( } #[tokio::test] -#[serial] async fn test_set_remote_target_rejects_unversioned_source_bucket() -> Result<(), Box> { init_logging(); @@ -2731,7 +2725,6 @@ async fn test_set_remote_target_rejects_unversioned_source_bucket() -> Result<() } #[tokio::test] -#[serial] async fn test_replication_check_rejects_unversioned_source_bucket() -> Result<(), Box> { init_logging(); @@ -2755,7 +2748,6 @@ async fn test_replication_check_rejects_unversioned_source_bucket() -> Result<() } #[tokio::test] -#[serial] async fn test_replication_check_rejects_missing_replication_config() -> Result<(), Box> { init_logging(); @@ -2779,7 +2771,6 @@ async fn test_replication_check_rejects_missing_replication_config() -> Result<( } #[tokio::test] -#[serial] async fn test_replication_check_rejects_invalid_bucket() -> Result<(), Box> { init_logging(); @@ -2798,7 +2789,6 @@ async fn test_replication_check_rejects_invalid_bucket() -> Result<(), Box Result<(), Box> { init_logging(); @@ -2842,7 +2832,6 @@ async fn test_set_remote_target_rejects_same_bucket_on_same_deployment() -> Resu } #[tokio::test] -#[serial] async fn test_set_remote_target_rejects_unversioned_target_bucket() -> Result<(), Box> { init_logging(); @@ -2876,7 +2865,6 @@ async fn test_set_remote_target_rejects_unversioned_target_bucket() -> Result<() } #[tokio::test] -#[serial] async fn test_set_remote_target_update_requires_arn() -> Result<(), Box> { init_logging(); @@ -2928,7 +2916,6 @@ async fn test_set_remote_target_update_requires_arn() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3017,7 +3004,6 @@ async fn fetch_single_target( } #[tokio::test] -#[serial] async fn test_set_remote_target_partial_update_preserves_credentials() -> Result<(), Box> { init_logging(); @@ -3103,7 +3089,6 @@ async fn test_set_remote_target_partial_update_preserves_credentials() -> Result } #[tokio::test] -#[serial] async fn test_set_remote_target_rejects_invalid_target_url() -> Result<(), Box> { init_logging(); @@ -3145,7 +3130,6 @@ async fn test_set_remote_target_rejects_invalid_target_url() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3230,7 +3214,6 @@ async fn test_set_remote_target_rejects_self_signed_https_target_without_skip_tl } #[tokio::test] -#[serial] async fn test_set_remote_target_allows_self_signed_https_target_with_skip_tls_verify() -> Result<(), Box> { init_logging(); @@ -3342,7 +3325,6 @@ async fn test_set_remote_target_allows_self_signed_https_target_with_skip_tls_ve } #[tokio::test] -#[serial] async fn test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_pem() -> Result<(), Box> { init_logging(); @@ -3427,7 +3409,6 @@ async fn test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_ } #[tokio::test] -#[serial] async fn test_set_remote_target_allows_private_ca_https_target_with_ca_cert_pem() -> Result<(), Box> { init_logging(); @@ -3518,7 +3499,6 @@ async fn test_set_remote_target_allows_private_ca_https_target_with_ca_cert_pem( } #[tokio::test] -#[serial] async fn test_list_remote_targets_rejects_empty_bucket() -> Result<(), Box> { init_logging(); @@ -3538,7 +3518,6 @@ async fn test_list_remote_targets_rejects_empty_bucket() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3557,7 +3536,6 @@ async fn test_list_remote_targets_rejects_invalid_bucket() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3598,7 +3576,6 @@ async fn test_remove_remote_target_rejects_missing_target() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3623,7 +3600,6 @@ async fn test_remove_remote_target_rejects_missing_arn() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3647,7 +3623,6 @@ async fn test_remove_remote_target_rejects_invalid_bucket() -> Result<(), Box Result<(), Box> { init_logging(); @@ -3687,7 +3662,6 @@ async fn test_remove_remote_target_rejects_target_used_by_replication() -> Resul } #[tokio::test] -#[serial] async fn test_delete_bucket_replication_removes_remote_target() -> Result<(), Box> { init_logging(); @@ -3737,7 +3711,6 @@ async fn test_delete_bucket_replication_removes_remote_target() -> Result<(), Bo } #[tokio::test] -#[serial] async fn test_bucket_replication_replicates_put_object_issue_2539() -> Result<(), Box> { init_logging(); @@ -3779,7 +3752,6 @@ async fn test_bucket_replication_replicates_put_object_issue_2539() -> Result<() } #[tokio::test] -#[serial] async fn test_bucket_replication_converges_delete_marker_and_version_purge() -> TestResult { init_logging(); @@ -3878,7 +3850,6 @@ async fn test_bucket_replication_converges_delete_marker_and_version_purge() -> } #[tokio::test] -#[serial] async fn test_bucket_replication_disabled_delete_marker_does_not_propagate() -> TestResult { init_logging(); @@ -3965,7 +3936,6 @@ async fn test_bucket_replication_disabled_delete_marker_does_not_propagate() -> /// interoperability profile for a runner that provisions MinIO credentials /// and a reachable endpoint. #[tokio::test] -#[serial] async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestResult { init_logging(); @@ -4293,7 +4263,6 @@ async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestR } #[tokio::test] -#[serial] async fn test_single_bucket_multipart_replication_fans_out_to_multiple_targets() -> Result<(), Box> { init_logging(); @@ -4462,7 +4431,6 @@ async fn test_repl17_failure_observation_helpers() -> TestResult { /// the replica is decryptable only with the original customer key. The /// backlog#1291 property still holds: never a silent plaintext replica. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_c_contract() -> TestResult { init_logging(); @@ -4540,7 +4508,6 @@ async fn test_bucket_replication_sse_c_contract() -> TestResult { /// part — part boundaries and the encrypted-multipart marker survive so the /// replica decrypts each part with its part-derived nonce. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_c_multipart_passthrough() -> TestResult { init_logging(); @@ -4657,7 +4624,6 @@ async fn test_bucket_replication_sse_c_multipart_passthrough() -> TestResult { /// (independent KMS, so success proves target-owned envelopes), preserved /// source ETag, and a version that stays stable across scanner cycles. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_s3_contract() -> TestResult { init_logging(); assert_managed_sse_replicates_and_reencrypts("sse-s3", false).await @@ -4667,7 +4633,6 @@ async fn test_bucket_replication_sse_s3_contract() -> TestResult { /// fail closed — replication FAILED, and no plaintext (or any) replica ever /// materializes on the target. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_s3_fails_closed_without_target_kms() -> TestResult { init_logging(); @@ -4711,7 +4676,6 @@ async fn test_bucket_replication_sse_s3_fails_closed_without_target_kms() -> Tes /// the ETag comparison sees the preserved source ETag on the replica and does /// not rewrite it, so the replica's version stays stable through the resync. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_s3_resync_converges() -> TestResult { init_logging(); @@ -4768,7 +4732,6 @@ async fn test_bucket_replication_sse_s3_resync_converges() -> TestResult { /// re-encrypts under its own default key. The independent-KMS pair proves the /// replica's envelope is target-owned. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_kms_contract() -> TestResult { init_logging(); assert_managed_sse_replicates_and_reencrypts("sse-kms", true).await @@ -4779,7 +4742,6 @@ async fn test_bucket_replication_sse_kms_contract() -> TestResult { /// carries the full header set (SSE intent, content-type, user metadata) and /// the completed replica preserves the source's multipart ETag. #[tokio::test] -#[serial] async fn test_bucket_replication_sse_s3_multipart_reencrypts() -> TestResult { init_logging(); @@ -4873,7 +4835,6 @@ async fn test_bucket_replication_sse_s3_multipart_reencrypts() -> TestResult { /// still-running source's data scanner (short cycle via [`FAST_SCANNER_ENV`]) /// re-drives the failed objects once the target is reachable again. #[tokio::test] -#[serial] async fn test_bucket_replication_recovers_after_target_outage() -> TestResult { init_logging(); @@ -4953,7 +4914,6 @@ async fn test_bucket_replication_recovers_after_target_outage() -> TestResult { /// must settle back to zero even though the historical failed counter remains /// non-zero. #[tokio::test] -#[serial] async fn test_bucket_replication_backlog_metrics_observe_outage_and_recovery() -> TestResult { init_logging(); @@ -5087,7 +5047,6 @@ async fn test_bucket_replication_backlog_metrics_observe_outage_and_recovery() - /// must converge every persisted failure, including the replayed delete marker /// (whose replication decision is re-derived from the live config). #[tokio::test] -#[serial] async fn test_bucket_replication_replays_failed_entries_after_source_restart() -> TestResult { init_logging(); @@ -5179,7 +5138,6 @@ async fn test_bucket_replication_replays_failed_entries_after_source_restart() - } #[tokio::test] -#[serial] async fn test_bucket_replication_replayed_delete_marker_preserves_source_mtime_without_source_restart() -> TestResult { init_logging(); @@ -5249,7 +5207,6 @@ async fn test_bucket_replication_replayed_delete_marker_preserves_source_mtime_w } #[tokio::test] -#[serial] async fn test_sequential_bucket_replication_succeeds_for_multiple_buckets() -> Result<(), Box> { init_logging(); @@ -5293,7 +5250,6 @@ async fn test_sequential_bucket_replication_succeeds_for_multiple_buckets() -> R } #[tokio::test] -#[serial] async fn test_replication_recovers_after_runtime_target_cache_is_cleared() -> Result<(), Box> { init_logging(); @@ -5337,7 +5293,6 @@ async fn test_replication_recovers_after_runtime_target_cache_is_cleared() -> Re } #[tokio::test] -#[serial] async fn test_site_replication_allows_self_signed_https_with_skip_tls_verify_real_dual_node() -> TestResult { init_logging(); @@ -5416,7 +5371,6 @@ async fn test_site_replication_allows_self_signed_https_with_skip_tls_verify_rea } #[tokio::test] -#[serial] async fn test_site_replication_allows_private_ca_https_with_ca_cert_pem_real_dual_node() -> TestResult { init_logging(); @@ -5495,7 +5449,6 @@ async fn test_site_replication_allows_private_ca_https_with_ca_cert_pem_real_dua } #[tokio::test] -#[serial] async fn test_site_replication_resync_lifecycle_survives_real_server_restart() -> Result<(), Box> { init_logging(); let resync_process_env = [ @@ -5715,7 +5668,6 @@ async fn test_site_replication_resync_lifecycle_survives_real_server_restart() - } #[tokio::test] -#[serial] async fn test_site_replication_edit_and_status_peer_state_real_three_node() -> Result<(), Box> { init_logging(); @@ -5964,7 +5916,6 @@ async fn test_site_replication_edit_and_status_peer_state_real_three_node() -> R } #[tokio::test] -#[serial] async fn test_site_replication_remove_all_real_dual_node() -> Result<(), Box> { init_logging(); @@ -6084,7 +6035,6 @@ async fn test_site_replication_remove_all_real_dual_node() -> Result<(), Box Result<(), Box> { init_logging(); @@ -6193,7 +6143,6 @@ async fn test_site_replication_state_edit_fresh_and_stale_real_dual_node() -> Re } #[tokio::test] -#[serial] async fn test_site_replication_replicates_object_with_bucket_versioning_real_dual_node() -> TestResult { init_logging(); @@ -6284,7 +6233,6 @@ async fn test_site_replication_replicates_object_with_bucket_versioning_real_dua /// receiver was dropped with only a debug line, while `replicate status` still reported /// "1/1 Buckets in sync" because both configs were byte-identical. #[tokio::test] -#[serial] async fn test_site_replication_config_broadcast_keeps_reverse_direction_real_dual_node() -> TestResult { init_logging(); @@ -6423,7 +6371,6 @@ async fn wait_for_site_replication_rule( } #[tokio::test] -#[serial] async fn test_site_replication_active_active_converges_without_loops_real_dual_node() -> TestResult { init_logging(); @@ -6741,7 +6688,6 @@ async fn test_site_replication_active_active_converges_without_loops_real_dual_n } #[tokio::test] -#[serial] async fn test_site_replication_replicates_policy_backed_user_access_real_dual_node() -> Result<(), Box> { init_logging(); @@ -6829,7 +6775,6 @@ async fn test_site_replication_replicates_policy_backed_user_access_real_dual_no } #[tokio::test] -#[serial] async fn test_site_replication_replicates_group_policy_backed_access_real_dual_node() -> Result<(), Box> { init_logging(); @@ -6920,7 +6865,6 @@ async fn test_site_replication_replicates_group_policy_backed_access_real_dual_n } #[tokio::test] -#[serial] async fn test_service_account_policy_from_accountinfo_round_trips_real_single_node() -> TestResult { init_logging(); @@ -6972,7 +6916,6 @@ async fn test_service_account_policy_from_accountinfo_round_trips_real_single_no } #[tokio::test] -#[serial] async fn test_site_replication_replicates_multiple_service_accounts_real_dual_node() -> Result<(), Box> { init_logging(); @@ -7073,7 +7016,6 @@ async fn test_site_replication_replicates_multiple_service_accounts_real_dual_no } #[tokio::test] -#[serial] async fn test_site_replication_replicates_service_accounts_created_from_sts_session_real_dual_node() -> TestResult { init_logging(); @@ -7214,7 +7156,6 @@ async fn wait_for_target_request_version_id( } #[tokio::test] -#[serial] async fn test_bucket_resync_restart_revisits_objects_before_out_of_order_checkpoint() -> TestResult { init_logging(); @@ -7333,7 +7274,6 @@ async fn test_bucket_resync_restart_revisits_objects_before_out_of_order_checkpo /// CreateMultipartUpload (the version is decided at initiate time) must both /// carry the source version as `?versionId=`. #[tokio::test] -#[serial] async fn test_replication_put_and_create_multipart_carry_source_version_id_query() -> TestResult { init_logging(); @@ -7448,7 +7388,6 @@ async fn test_replication_put_and_create_multipart_carry_source_version_id_query /// flow to the onward bucket, proving B's outbound replication and scanner /// are live. #[tokio::test] -#[serial] async fn test_scanner_never_cascades_inbound_replicas() -> TestResult { init_logging(); @@ -7529,7 +7468,6 @@ async fn test_scanner_never_cascades_inbound_replicas() -> TestResult { /// version ids and still mint its own there — the check must not report OK /// while multipart deletes and heals would silently miss. #[tokio::test] -#[serial] async fn test_replication_check_flags_multipart_only_version_minting_target() -> TestResult { init_logging(); @@ -7602,7 +7540,6 @@ async fn test_replication_check_flags_multipart_only_version_minting_target() -> } #[tokio::test] -#[serial] async fn test_replication_check_aborts_failed_multipart_probes() -> TestResult { init_logging(); @@ -7765,7 +7702,6 @@ async fn test_replication_check_aborts_failed_multipart_probes() -> TestResult { /// BucketRemoteTargetVersionMismatch — while still cleaning up the probe /// object via the version id the target actually assigned. #[tokio::test] -#[serial] async fn test_replication_check_flags_version_minting_target() -> TestResult { init_logging(); @@ -8006,7 +7942,6 @@ async fn wait_for_target_marker_purged( /// the target forever. Contract under test: a failed purge attempt is retried /// within the watch window and converges once the fault clears. #[tokio::test] -#[serial] async fn test_delayed_delete_marker_purge_retries_after_transient_target_failure() -> TestResult { init_logging(); let source_bucket = "delayed-purge-retry-src"; @@ -8066,7 +8001,6 @@ async fn test_delayed_delete_marker_purge_retries_after_transient_target_failure /// with an idempotent 204, which used to look like success and strand the /// real marker on the target forever. #[tokio::test] -#[serial] async fn test_delayed_delete_marker_purge_uses_target_assigned_version() -> TestResult { init_logging(); let source_bucket = "delayed-purge-mint-src"; @@ -8099,7 +8033,6 @@ async fn test_delayed_delete_marker_purge_uses_target_assigned_version() -> Test /// replayed purge succeeds, the entry must be acknowledged instead of being /// retained as Missed forever. #[tokio::test] -#[serial] async fn test_delayed_delete_marker_purge_exhaustion_persists_to_mrf_and_replays_on_restart() -> TestResult { init_logging(); let source_bucket = "delayed-purge-mrf-src"; @@ -8236,7 +8169,6 @@ async fn build_scanner_compensation_pair( /// nil-version objects entirely (`scanner_folder.rs` heal_replication), so it /// must NEVER be compensated. #[tokio::test] -#[serial] async fn test_scanner_compensates_existing_objects_across_write_paths() -> TestResult { init_logging(); let source_bucket = "scanner-comp-src"; @@ -8352,7 +8284,6 @@ async fn test_scanner_compensates_existing_objects_across_write_paths() -> TestR /// written after the rule replicate normally (the setting only gates the /// existing-object resync path). #[tokio::test] -#[serial] async fn test_scanner_never_compensates_when_existing_object_replication_disabled() -> TestResult { init_logging(); let source_bucket = "scanner-disabled-src"; From 382ae9529ec447e10010d11cbb4c095cf0e1d237 Mon Sep 17 00:00:00 2001 From: houseme Date: Tue, 18 Aug 2026 17:20:55 +0800 Subject: [PATCH 13/24] feat(ecstore): instrument rename sync tail metrics (#6205) Add default-off PUT stage attribution for the rename_data sync tail so strict durability probes can split queue wait, fdatasync, directory fsync, rename, per-disk wait, and quorum wait without changing commit ordering or S3-visible behavior. Co-authored-by: heihutu --- crates/ecstore/src/disk/local.rs | 82 +++++++++++--- crates/ecstore/src/disk/os.rs | 101 ++++++++++++++++-- .../src/set_disk/core/io_primitives.rs | 19 +++- crates/io-metrics/src/lib.rs | 69 ++++++++++++ 4 files changed, 246 insertions(+), 25 deletions(-) diff --git a/crates/ecstore/src/disk/local.rs b/crates/ecstore/src/disk/local.rs index 1932744e8..a71f2c55b 100644 --- a/crates/ecstore/src/disk/local.rs +++ b/crates/ecstore/src/disk/local.rs @@ -9264,17 +9264,27 @@ impl DiskAPI for LocalDisk { // accept that window (documented in docs/operations/durability-modes.md). if durability.syncs_commit_metadata() && let Some(parent) = dst_file_path.parent() - && let Err(err) = os::fsync_dir(parent).await { - rollback_committed_rename_std(&dst_file_path, committed_new_data_path, rollback_data_dir) - .map_err(to_file_error)?; - // The commit rename changed the dst part inodes before this fsync - // failed and rolled them back; drop any fd cached during that - // window so readers re-open the restored inode (rustfs/backlog#1177). - for part_path in &invalidate_part_paths { - self.io_backend.invalidate_cached_fd(dst_volume, part_path).await; + let fsync_started = rustfs_io_metrics::put_stage_timer(); + if let Err(err) = os::fsync_dir(parent).await { + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC, + fsync_started, + ); + rollback_committed_rename_std(&dst_file_path, committed_new_data_path, rollback_data_dir) + .map_err(to_file_error)?; + // The commit rename changed the dst part inodes before this fsync + // failed and rolled them back; drop any fd cached during that + // window so readers re-open the restored inode (rustfs/backlog#1177). + for part_path in &invalidate_part_paths { + self.io_backend.invalidate_cached_fd(dst_volume, part_path).await; + } + return Err(to_file_error(err).into()); } - return Err(to_file_error(err).into()); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC, + fsync_started, + ); } // First PUT of an object creates its directory (and any missing prefix @@ -9293,7 +9303,12 @@ impl DiskAPI for LocalDisk { if !dir.starts_with(&dst_volume_dir) { break; } + let fsync_started = rustfs_io_metrics::put_stage_timer(); if let Err(err) = os::fsync_dir(dir).await { + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC, + fsync_started, + ); rollback_committed_rename_std(&dst_file_path, committed_new_data_path, rollback_data_dir) .map_err(to_file_error)?; // Same post-commit rollback window as above — drop cached @@ -9304,6 +9319,10 @@ impl DiskAPI for LocalDisk { } return Err(to_file_error(err).into()); } + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC, + fsync_started, + ); if dir == dst_volume_dir.as_path() { break; } @@ -9532,10 +9551,21 @@ impl DiskAPI for LocalDisk { } if let Some(admission) = file_sync_admission.as_ref() && let Some(backup_parent) = backup_path.parent() - && let Err(err) = - os::fsync_dir_with_namespace_file_sync_limit(backup_parent, mutation_lease.clone(), admission).await { - return Err(DiskError::from(to_file_error(err))); + let fsync_started = rustfs_io_metrics::put_stage_timer(); + if let Err(err) = + os::fsync_dir_with_namespace_file_sync_limit(backup_parent, mutation_lease.clone(), admission).await + { + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_BACKUP_DIR_FSYNC, + fsync_started, + ); + return Err(DiskError::from(to_file_error(err))); + } + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_BACKUP_DIR_FSYNC, + fsync_started, + ); } local_rollback_path = None; } @@ -9573,11 +9603,22 @@ impl DiskAPI for LocalDisk { // Persist the commit rename's directory entry across power loss. if let Some(admission) = file_sync_admission.as_ref() && let Some(dst_parent) = dst_file_path.parent() - && let Err(err) = - os::fsync_dir_with_namespace_file_sync_limit(dst_parent, mutation_lease.clone(), admission).await { - rollback_inline_metadata_commit_std(&dst_file_path, rollback_data_dir, local_rollback_path.as_deref())?; - return Err(err); + let fsync_started = rustfs_io_metrics::put_stage_timer(); + if let Err(err) = + os::fsync_dir_with_namespace_file_sync_limit(dst_parent, mutation_lease.clone(), admission).await + { + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC, + fsync_started, + ); + rollback_inline_metadata_commit_std(&dst_file_path, rollback_data_dir, local_rollback_path.as_deref())?; + return Err(err); + } + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC, + fsync_started, + ); } // Same power-loss gap as the non-inline path (rustfs/backlog#922 @@ -9595,9 +9636,14 @@ impl DiskAPI for LocalDisk { if !ancestor_dir.starts_with(&dst_volume_dir) { break; } + let fsync_started = rustfs_io_metrics::put_stage_timer(); if let Err(err) = os::fsync_dir_with_namespace_file_sync_limit(ancestor_dir, mutation_lease.clone(), admission).await { + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC, + fsync_started, + ); rollback_inline_metadata_commit_std( &dst_file_path, rollback_data_dir, @@ -9605,6 +9651,10 @@ impl DiskAPI for LocalDisk { )?; return Err(err); } + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC, + fsync_started, + ); if ancestor_dir == dst_volume_dir.as_path() { break; } diff --git a/crates/ecstore/src/disk/os.rs b/crates/ecstore/src/disk/os.rs index f5b1c5f66..decb36964 100644 --- a/crates/ecstore/src/disk/os.rs +++ b/crates/ecstore/src/disk/os.rs @@ -343,6 +343,7 @@ pub(crate) async fn acquire_rename_data_mutation_lease( /// this order uniform prevents one slow disk from reserving global capacity /// while it waits for its own concurrency slot. async fn acquire_file_sync_permits(disk_permits: Arc) -> io::Result<(OwnedSemaphorePermit, SemaphorePermit<'static>)> { + let wait_started = rustfs_io_metrics::put_stage_timer(); let disk_permit = disk_permits .acquire_owned() .await @@ -351,6 +352,10 @@ async fn acquire_file_sync_permits(disk_permits: Arc) -> io::Result<( .acquire() .await .map_err(|_| io::Error::other("global file sync concurrency limiter closed"))?; + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_FILE_SYNC_PERMIT_WAIT, + wait_started, + ); Ok((disk_permit, global_permit)) } @@ -551,9 +556,19 @@ pub(crate) fn sync_file(path: &Path) -> io::Result<()> { file.sync_data() } +fn sync_file_with_put_stage_metric(path: &Path) -> io::Result<()> { + let sync_started = rustfs_io_metrics::put_stage_timer(); + let result = sync_file(path); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_FILE_FDATASYNC, + sync_started, + ); + result +} + fn sync_files(paths: &[PathBuf]) -> io::Result<()> { for path in paths { - sync_file(path)?; + sync_file_with_put_stage_metric(path)?; } Ok(()) } @@ -599,7 +614,13 @@ pub(crate) async fn sync_dir_files_with_limiter(dir: impl AsRef, disk_perm let files = regular_files(&scan_dir)?; if files.len() < PARALLEL_FILE_SYNC_THRESHOLD { sync_files(&files)?; - fsync_dir_std(scan_dir)?; + let fsync_started = rustfs_io_metrics::put_stage_timer(); + let result = fsync_dir_std(scan_dir); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_SRC_DIR_FSYNC, + fsync_started, + ); + result?; return Ok(None); } Ok::<_, io::Error>(Some(files)) @@ -612,10 +633,19 @@ pub(crate) async fn sync_dir_files_with_limiter(dir: impl AsRef, disk_perm futures::stream::iter(files.into_iter().map(Ok::<_, io::Error>)) .try_for_each_concurrent(MAX_PARALLEL_FILE_SYNCS, |path| { let disk_permits = disk_permits.clone(); - async move { run_file_sync_blocking(disk_permits, move || sync_file(&path)).await } + async move { run_file_sync_blocking(disk_permits, move || sync_file_with_put_stage_metric(&path)).await } }) .await?; - run_file_sync_blocking(disk_permits, move || fsync_dir_std(dir)).await + run_file_sync_blocking(disk_permits, move || { + let fsync_started = rustfs_io_metrics::put_stage_timer(); + let result = fsync_dir_std(dir); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_SRC_DIR_FSYNC, + fsync_started, + ); + result + }) + .await } /// Check if the given disk path is the root disk. @@ -1174,10 +1204,15 @@ pub(crate) struct FileSyncAdmission { } pub(crate) async fn acquire_file_sync_admission(disk_permits: Arc) -> io::Result { + let wait_started = rustfs_io_metrics::put_stage_timer(); let disk_permit = disk_permits .acquire_owned() .await .map_err(|_| io::Error::other("disk file sync concurrency limiter closed"))?; + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_FILE_SYNC_PERMIT_WAIT, + wait_started, + ); Ok(FileSyncAdmission { disk_permit: Arc::new(disk_permit), }) @@ -1200,10 +1235,15 @@ async fn run_blocking_namespace_file_sync_operation_with_global io::Result + Send + 'static, ) -> io::Result { + let wait_started = rustfs_io_metrics::put_stage_timer(); let global_permit = global_permits .acquire() .await .map_err(|_| io::Error::other("global file sync concurrency limiter closed"))?; + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_GLOBAL_FILE_SYNC_PERMIT_WAIT, + wait_started, + ); let disk_permit = admission.disk_permit.clone(); let result = tokio::task::spawn_blocking(move || { let _lease = lease; @@ -1420,7 +1460,13 @@ fn rename_into_existing_parent( use rustix::fs::{Mode, OFlags, open, renameat}; let Some(parent_guard) = parent_guard else { - return super::fs::rename_std(src_file_path, dst_file_path); + let rename_started = rustfs_io_metrics::put_stage_timer(); + let result = super::fs::rename_std(src_file_path, dst_file_path); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_RENAME_SYSCALL, + rename_started, + ); + return result; }; let src_parent = src_file_path .parent() @@ -1441,7 +1487,13 @@ fn rename_into_existing_parent( .last() .ok_or_else(|| io::Error::other("rename destination parent guard is empty"))?; - renameat(&src_parent, src_name, dst_parent, dst_name).map_err(io::Error::from) + let rename_started = rustfs_io_metrics::put_stage_timer(); + let result = renameat(&src_parent, src_name, dst_parent, dst_name).map_err(io::Error::from); + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_RENAME_SYSCALL, + rename_started, + ); + result } #[cfg(windows)] @@ -2890,6 +2942,7 @@ pub fn is_dir_not_empty_error(err: &io::Error) -> bool { #[cfg(test)] mod tests { use super::*; + use crate::test_metrics::CapturingRecorder; use std::sync::Mutex; use std::time::Duration; use tempfile::tempdir; @@ -2910,6 +2963,42 @@ mod tests { PublicationRoot::new(&common).expect("test publication root should open") } + #[test] + #[serial_test::serial(file_sync_metrics)] + fn sync_file_with_put_stage_metric_records_fdatasync_only_when_enabled() { + let previous_gate = rustfs_io_metrics::put_stage_metrics_enabled(); + rustfs_io_metrics::set_put_stage_metrics_enabled(false); + let dir = tempdir().expect("temp dir should be created"); + let path = dir.path().join("part.1"); + std::fs::write(&path, b"payload").expect("test file should be written"); + let recorder = CapturingRecorder::default(); + + metrics::with_local_recorder(&recorder, || { + sync_file_with_put_stage_metric(&path).expect("disabled metric sync_file should succeed"); + assert_eq!( + recorder.histogram_sample_count("rustfs_s3_put_object_stage_duration_ms"), + 0, + "disabled PUT stage metrics must not emit fdatasync samples" + ); + + rustfs_io_metrics::set_put_stage_metrics_enabled(true); + sync_file_with_put_stage_metric(&path).expect("enabled metric sync_file should succeed"); + rustfs_io_metrics::set_put_stage_metrics_enabled(false); + }); + + assert_eq!( + recorder + .histogram_values( + "rustfs_s3_put_object_stage_duration_ms", + &[("stage", rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_FILE_FDATASYNC)] + ) + .len(), + 1, + "enabled PUT stage metrics must emit one fdatasync sample" + ); + rustfs_io_metrics::set_put_stage_metrics_enabled(previous_gate); + } + async fn rename_all( src_file_path: impl AsRef, dst_file_path: impl AsRef, diff --git a/crates/ecstore/src/set_disk/core/io_primitives.rs b/crates/ecstore/src/set_disk/core/io_primitives.rs index 7af25a64c..dcf5d11e3 100644 --- a/crates/ecstore/src/set_disk/core/io_primitives.rs +++ b/crates/ecstore/src/set_disk/core/io_primitives.rs @@ -3389,8 +3389,15 @@ impl SetDisks { // A no-op immediately-ready future in production. Self::rename_fanout_barrier(&dst_object, i, rename_fanout_barrier_phase::RENAME).await; - disk.rename_data_borrowed(&src_bucket, &src_object, file_info, &dst_bucket, &dst_object) - .await + let disk_wait_started = rustfs_io_metrics::put_stage_timer(); + let result = disk + .rename_data_borrowed(&src_bucket, &src_object, file_info, &dst_bucket, &dst_object) + .await; + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_DISK_WAIT, + disk_wait_started, + ); + result }) .catch_unwind() }); @@ -3403,7 +3410,13 @@ impl SetDisks { let mut cleanup_data_dirs = vec![None; disk_count]; let mut old_current_sizes = vec![None; disk_count]; - let (results, mut file_infos) = fanout.await.map_err(|_| DiskError::Unexpected)?; + let quorum_wait_started = rustfs_io_metrics::put_stage_timer(); + let fanout_result = fanout.await; + rustfs_io_metrics::record_put_object_stage_duration_from( + rustfs_io_metrics::PUT_STAGE_SET_DISK_RENAME_QUORUM_WAIT, + quorum_wait_started, + ); + let (results, mut file_infos) = fanout_result.map_err(|_| DiskError::Unexpected)?; for (idx, result) in results.iter().enumerate() { match result { diff --git a/crates/io-metrics/src/lib.rs b/crates/io-metrics/src/lib.rs index 3f021772d..86b38ed5a 100644 --- a/crates/io-metrics/src/lib.rs +++ b/crates/io-metrics/src/lib.rs @@ -109,6 +109,17 @@ pub fn put_stage_timer() -> Option { put_stage_metrics_enabled().then(std::time::Instant::now) } +pub const PUT_STAGE_SET_DISK_RENAME_QUORUM_WAIT: &str = "set_disk_rename_quorum_wait"; +pub const PUT_STAGE_SET_DISK_RENAME_DISK_WAIT: &str = "set_disk_rename_disk_wait"; +pub const PUT_STAGE_SET_DISK_RENAME_FILE_SYNC_PERMIT_WAIT: &str = "set_disk_rename_file_sync_permit_wait"; +pub const PUT_STAGE_SET_DISK_RENAME_GLOBAL_FILE_SYNC_PERMIT_WAIT: &str = "set_disk_rename_global_file_sync_permit_wait"; +pub const PUT_STAGE_SET_DISK_RENAME_FILE_FDATASYNC: &str = "set_disk_rename_file_fdatasync"; +pub const PUT_STAGE_SET_DISK_RENAME_SRC_DIR_FSYNC: &str = "set_disk_rename_src_dir_fsync"; +pub const PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC: &str = "set_disk_rename_dst_dir_fsync"; +pub const PUT_STAGE_SET_DISK_RENAME_BACKUP_DIR_FSYNC: &str = "set_disk_rename_backup_dir_fsync"; +pub const PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC: &str = "set_disk_rename_ancestor_dir_fsync"; +pub const PUT_STAGE_SET_DISK_RENAME_RENAME_SYSCALL: &str = "set_disk_rename_rename_syscall"; + #[inline(always)] pub fn get_stage_metrics_enabled() -> bool { GET_STAGE_METRICS_ENABLED.load(Ordering::Relaxed) @@ -2618,6 +2629,7 @@ mod tests { use super::*; use metrics_util::MetricKind; use metrics_util::debugging::{DebugValue, DebuggingRecorder}; + use std::collections::HashSet; use std::sync::{Arc, Barrier, Mutex}; // Serialize tests that mutate the process-global PUT_STAGE_METRICS_ENABLED flag. @@ -2861,6 +2873,63 @@ mod tests { set_put_stage_metrics_enabled(false); } + #[test] + fn put_stage_sync_tail_labels_are_static_and_gated() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let stages = [ + PUT_STAGE_SET_DISK_RENAME_QUORUM_WAIT, + PUT_STAGE_SET_DISK_RENAME_DISK_WAIT, + PUT_STAGE_SET_DISK_RENAME_FILE_SYNC_PERMIT_WAIT, + PUT_STAGE_SET_DISK_RENAME_GLOBAL_FILE_SYNC_PERMIT_WAIT, + PUT_STAGE_SET_DISK_RENAME_FILE_FDATASYNC, + PUT_STAGE_SET_DISK_RENAME_SRC_DIR_FSYNC, + PUT_STAGE_SET_DISK_RENAME_DST_DIR_FSYNC, + PUT_STAGE_SET_DISK_RENAME_BACKUP_DIR_FSYNC, + PUT_STAGE_SET_DISK_RENAME_ANCESTOR_DIR_FSYNC, + PUT_STAGE_SET_DISK_RENAME_RENAME_SYSCALL, + ]; + let unique = stages.iter().copied().collect::>(); + assert_eq!(unique.len(), stages.len()); + assert!( + stages + .iter() + .all(|stage| stage.starts_with("set_disk_rename_") && !stage.contains('/') && !stage.contains('{')) + ); + + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); + metrics::with_local_recorder(&recorder, || { + set_put_stage_metrics_enabled(false); + for stage in stages { + record_put_object_stage_duration(stage, 1.0); + } + set_put_stage_metrics_enabled(true); + for stage in stages { + record_put_object_stage_duration(stage, 1.0); + } + set_put_stage_metrics_enabled(false); + }); + + let recorded = snapshotter + .snapshot() + .into_vec() + .into_iter() + .filter(|(composite, _, _, _)| { + composite.kind() == MetricKind::Histogram && composite.key().name() == "rustfs_s3_put_object_stage_duration_ms" + }) + .flat_map(|(composite, _, _, _)| { + composite + .key() + .labels() + .filter(|label| label.key() == "stage") + .map(|label| label.value().to_string()) + .collect::>() + }) + .collect::>(); + assert_eq!(recorded.len(), stages.len()); + assert!(stages.iter().all(|stage| recorded.contains(*stage))); + } + #[test] fn test_put_object_diagnostic_buckets() { assert_eq!(put_object_size_bucket(0), "unknown"); From a38743caf5a54c18a6c99bbc73d5ccbe2bc42c2b Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 18:27:01 +0800 Subject: [PATCH 14/24] chore(zip): trim the unused extract and create surface (#6210) rustfs-zip has one workspace consumer, and it uses only CompressionFormat::{from_extension, extension, get_decoder} and ArchiveLimits. Remove the tar/zip extract, zip create, and in-memory compress helpers together with the types and dependencies that only served them. Trimming public API is semver-major once the stable tag is cut, so it costs least now. --- Cargo.lock | 5 - crates/zip/Cargo.toml | 16 +- crates/zip/README.md | 54 +- crates/zip/benches/zip_benchmark.rs | 416 ------- crates/zip/src/lib.rs | 1685 +-------------------------- 5 files changed, 58 insertions(+), 2118 deletions(-) delete mode 100644 crates/zip/benches/zip_benchmark.rs diff --git a/Cargo.lock b/Cargo.lock index 0fb09e78e..82f66641e 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -10489,15 +10489,10 @@ dependencies = [ name = "rustfs-zip" version = "1.0.0-rc.2" dependencies = [ - "astral-tokio-tar", "async-compression", - "criterion", "hotpath", - "tempfile", "thiserror 2.0.20", "tokio", - "tokio-stream", - "zip", ] [[package]] diff --git a/crates/zip/Cargo.toml b/crates/zip/Cargo.toml index 54cdd9547..01ec5d84e 100644 --- a/crates/zip/Cargo.toml +++ b/crates/zip/Cargo.toml @@ -20,7 +20,7 @@ repository.workspace = true rust-version.workspace = true version.workspace = true homepage.workspace = true -description = "ZIP file handling for RustFS, providing support for reading and writing ZIP archives." +description = "Archive format detection and async stream decoders for RustFS." keywords = ["zip", "compression", "rustfs", "Minio"] categories = ["web-programming", "development-tools", "compression"] documentation = "https://docs.rs/rustfs-zip/latest/rustfs_zip/" @@ -28,10 +28,6 @@ documentation = "https://docs.rs/rustfs-zip/latest/rustfs_zip/" [lib] doctest = false -[[bench]] -name = "zip_benchmark" -harness = false - [features] default = [] hotpath = ["hotpath/hotpath", "hotpath/tokio"] @@ -48,16 +44,8 @@ async-compression = { workspace = true, features = [ "zstd", "xz", ] } -tokio = { workspace = true, features = ["fs", "io-util", "macros", "rt-multi-thread"] } -tokio-stream = { workspace = true } -astral-tokio-tar = { workspace = true } +tokio = { workspace = true, features = ["io-util", "macros", "rt"] } thiserror = { workspace = true } -zip = { workspace = true } - -[dev-dependencies] -criterion = { workspace = true, features = ["html_reports"] } -tempfile = { workspace = true } - [lints] workspace = true diff --git a/crates/zip/README.md b/crates/zip/README.md index 78e105f5f..a9b7865d5 100644 --- a/crates/zip/README.md +++ b/crates/zip/README.md @@ -1,9 +1,9 @@ [![RustFS](https://rustfs.com/images/rustfs-github.png)](https://rustfs.com) -# RustFS Zip - Archive And Compression Primitives +# RustFS Zip - Archive Format Detection And Stream Decoding

- High-performance compression and archiving for RustFS object storage + Archive format detection and async stream decoders for RustFS object storage

@@ -17,53 +17,23 @@ ## 📖 Overview -**RustFS Zip** provides archive and compression primitives for the [RustFS](https://rustfs.com) distributed object storage system. Today it is primarily used by RustFS archive extract flows to: +**RustFS Zip** provides the archive primitives used by the [RustFS](https://rustfs.com) archive extract flow: -- identify archive/compression formats by extension -- stream tar and tar+compression inputs through async decoders -- provide small ZIP read/write helpers for local archive workflows +- identify a compression format from an archive extension +- wrap an async reader in the matching stream decoder +- carry the shared default archive guardrails ## Current Features -- A clearer type model with: - - `CompressionCodec` for stream codecs - - `ArchiveKind` for container families - - `ArchiveFormat` for concrete archive/container combinations -- Async stream codecs for `gzip`, `bzip2`, `zlib`, `xz`, and `zstd` -- Tar archive iteration over async readers through `read_archive_entries()` / `extract_tar_entries()` -- Archive guardrails through `ArchiveLimits` for entry count, entry size, total unpacked size, and path length -- In-memory compression helpers for payload round-trip workflows -- Blocking ZIP create/extract helpers for local archive files -- ZIP helper metadata via `ZipEntry`, including: - - `compression_method` - - `archive_kind` - - `format` - - `unix_mode` -- ZIP helper options via `ZipWriteOptions`, including: - - `compression_level` - - `create_directory_entries` - -## Compatibility - -- `CompressionFormat` is retained as a compatibility layer for existing callers -- New code should prefer `ArchiveFormat`, `ArchiveKind`, and `CompressionCodec` when expressing archive semantics - -## ZIP Helper Scope - -The file-based ZIP helper APIs are best suited for: - -- local archive import/export flows -- admin-side packaging helpers -- test fixtures and tooling - -They are not intended to be a remote streaming ZIP access engine. +- `CompressionFormat::from_extension()` for extension-based format detection, including tar-family suffixes such as `tgz`, `tbz2`, `txz`, and `tzst` +- `CompressionFormat::get_decoder()` for async stream decoding of `gzip`, `bzip2`, `zlib`, `xz`, and `zstd`, plus a pass-through reader for plain `tar` +- `ArchiveLimits` with the default entry count, entry size, total unpacked size, and path length guardrails ## Current Boundaries -- ZIP is supported via file-based helper APIs, not the tar-family async stream APIs -- Tar-family stream APIs are intended for `tar`, `tar.gz`, `tar.bz2`, `tar.xz`, `tar.zst`, and similar compressed tar flows -- Default archive guardrails are intentionally conservative and do not replace higher-level RustFS object-path validation -- This crate does not currently implement a general-purpose parallel archive engine +- ZIP has no stream decoder: `get_decoder()` rejects `CompressionFormat::Zip`, because ZIP needs central-directory semantics that a forward-only stream cannot provide +- This crate detects formats and hands back decoders; archive iteration, entry writing, and extraction to disk belong to the caller +- `ArchiveLimits` carries the values only; enforcement and the resulting protocol error belong to the caller - Archive extraction safety policy remains the responsibility of the RustFS caller for object-store flows ## 📚 Documentation diff --git a/crates/zip/benches/zip_benchmark.rs b/crates/zip/benches/zip_benchmark.rs deleted file mode 100644 index c7047918b..000000000 --- a/crates/zip/benches/zip_benchmark.rs +++ /dev/null @@ -1,416 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -use criterion::{BenchmarkId, Criterion, Throughput, criterion_group, criterion_main}; -use rustfs_zip::{ - ArchiveLimits, CompressionFormat, CompressionLevel, ZipWriteOptions, create_zip_with_options, extract_tar_entries, - extract_zip_to_path_with_limits, extract_zip_with_limits, -}; -use std::hint::black_box; -use std::sync::Arc; -use std::sync::atomic::{AtomicUsize, Ordering}; -use tempfile::tempdir; -use tokio::runtime::Builder; -use tokio_tar::{Builder as TarBuilder, Header}; -use zip::ZipArchive; - -fn build_runtime() -> tokio::runtime::Runtime { - Builder::new_current_thread() - .enable_all() - .build() - .expect("build tokio runtime for rustfs-zip benchmarks") -} - -async fn build_tar_payload(entry_count: usize, payload_size: usize) -> Vec { - let sink = tokio::io::duplex(64 * 1024); - let (writer, mut reader) = sink; - let write_task = tokio::spawn(async move { - let mut builder = TarBuilder::new(writer); - let payload = vec![b'a'; payload_size]; - for index in 0..entry_count { - let mut header = Header::new_gnu(); - header.set_size(payload.len() as u64); - header.set_mode(0o644); - header.set_cksum(); - builder - .append_data(&mut header, format!("entry-{index}.txt"), &payload[..]) - .await - .expect("append tar benchmark entry"); - } - builder.finish().await.expect("finish tar benchmark archive"); - }); - - let mut output = Vec::new(); - tokio::io::copy(&mut reader, &mut output) - .await - .expect("read tar benchmark archive"); - write_task.await.expect("join tar writer task"); - output -} - -async fn build_compressed_tar_payload(format: CompressionFormat, entry_count: usize, payload_size: usize) -> Vec { - let tar_payload = build_tar_payload(entry_count, payload_size).await; - rustfs_zip::Compressor::new(format) - .compress(&tar_payload) - .await - .expect("compress tar benchmark payload") -} - -fn bench_tar_family_extract(c: &mut Criterion) { - let runtime = build_runtime(); - let mut group = c.benchmark_group("zip_tar_family_extract"); - - for (name, format, entry_count, payload_size) in [ - ("tar_gzip_small_many", CompressionFormat::Gzip, 64usize, 256usize), - ("tar_zstd_medium", CompressionFormat::Zstd, 16usize, 16 * 1024usize), - ] { - let payload = runtime.block_on(build_compressed_tar_payload(format, entry_count, payload_size)); - group.throughput(Throughput::Bytes(payload.len() as u64)); - group.bench_with_input(BenchmarkId::new(name, payload.len()), &payload, |b, payload| { - b.iter(|| { - runtime.block_on(async { - let seen = Arc::new(AtomicUsize::new(0)); - let seen_ref = Arc::clone(&seen); - extract_tar_entries(std::io::Cursor::new(payload.clone()), format, move |_entry| { - let seen_ref = Arc::clone(&seen_ref); - async move { - seen_ref.fetch_add(1, Ordering::Relaxed); - Ok(()) - } - }) - .await - .expect("extract tar benchmark payload"); - black_box(seen.load(Ordering::Relaxed)); - }); - }); - }); - } - - group.finish(); -} - -fn bench_zip_helper_round_trip(c: &mut Criterion) { - let runtime = build_runtime(); - let mut group = c.benchmark_group("zip_helper_round_trip"); - - let zip_matrix = [ - ("stored_flat_32x128", CompressionLevel::Fastest, 32usize, 128usize, "flat"), - ("stored_nested_32x256", CompressionLevel::Fastest, 32usize, 256usize, "nested"), - ("stored_flat_256x128", CompressionLevel::Fastest, 256usize, 128usize, "flat"), - ("deflated_flat_32x1k", CompressionLevel::Best, 32usize, 1024usize, "flat"), - ("deflated_nested_256x1k", CompressionLevel::Best, 256usize, 1024usize, "nested"), - ("deflated_deep_1024x4k", CompressionLevel::Best, 1024usize, 4 * 1024usize, "deep"), - ]; - - for (name, compression_level, file_count, payload_size, layout) in zip_matrix { - let files = (0..file_count) - .map(|index| { - let path = match layout { - "flat" => format!("file-{index}.txt"), - "nested" => format!("batch-{}/file-{index}.txt", index % 8), - "deep" => format!("lvl1/lvl2-{}/lvl3-{}/file-{index}.txt", index % 16, index % 32), - _ => format!("file-{index}.txt"), - }; - (path, vec![b'b'; payload_size]) - }) - .collect::>(); - let total_bytes = (file_count * payload_size) as u64; - group.throughput(Throughput::Bytes(total_bytes)); - - group.bench_with_input(BenchmarkId::new(name, total_bytes), &files, |b, files| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - runtime.block_on(async { - create_zip_with_options( - &zip_path, - files.clone(), - ZipWriteOptions { - compression_level, - create_directory_entries: true, - }, - ) - .await - .expect("create zip benchmark archive"); - - let entries = extract_zip_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("extract zip benchmark archive"); - black_box(entries.len()); - }); - }); - }); - } - - group.finish(); -} - -fn bench_zip_helper_hotspot_breakdown(c: &mut Criterion) { - let runtime = build_runtime(); - let mut group = c.benchmark_group("zip_helper_hotspot_breakdown"); - let files = (0..32) - .map(|index| (format!("batch/file-{index}.txt"), vec![b'c'; 256])) - .collect::>(); - let total_bytes = (32 * 256) as u64; - group.throughput(Throughput::Bytes(total_bytes)); - - group.bench_function("fs_setup_cleanup_only", |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - black_box((zip_path, extract_path)); - }); - }); - - group.bench_function("zip_create_only_stored_small", |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - runtime.block_on(async { - create_zip_with_options( - &zip_path, - files.clone(), - ZipWriteOptions { - compression_level: CompressionLevel::Fastest, - create_directory_entries: true, - }, - ) - .await - .expect("create zip benchmark archive"); - }); - }); - }); - - let payload_for_extract = { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - runtime.block_on(async { - create_zip_with_options( - &zip_path, - files.clone(), - ZipWriteOptions { - compression_level: CompressionLevel::Fastest, - create_directory_entries: true, - }, - ) - .await - .expect("prepare zip benchmark extract payload"); - }); - std::fs::read(&zip_path).expect("read benchmark zip payload") - }; - - group.bench_function("zip_extract_only_stored_small", |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - std::fs::write(&zip_path, &payload_for_extract).expect("write benchmark zip payload"); - runtime.block_on(async { - let entries = extract_zip_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("extract zip benchmark archive"); - black_box(entries.len()); - }); - }); - }); - - group.bench_function("zip_extract_only_stored_small_summary_only", |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - std::fs::write(&zip_path, &payload_for_extract).expect("write benchmark zip payload"); - runtime.block_on(async { - let summary = extract_zip_to_path_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("extract zip benchmark summary path"); - black_box(summary.entry_count); - }); - }); - }); - - group.bench_function("zip_reader_only_stored_small", |b| { - b.iter(|| { - let cursor = std::io::Cursor::new(payload_for_extract.clone()); - let mut archive = ZipArchive::new(cursor).expect("open zip archive for reader-only benchmark"); - let mut total_bytes = 0usize; - for index in 0..archive.len() { - let mut zip_file = archive.by_index(index).expect("access zip entry by index"); - let enclosed_name = zip_file - .enclosed_name() - .expect("resolve enclosed zip entry name") - .to_string_lossy() - .replace('\\', "/"); - let size = zip_file.size(); - assert!(!enclosed_name.is_empty(), "zip reader-only benchmark expects non-empty names"); - assert!( - size <= ArchiveLimits::default().max_entry_size, - "zip reader-only benchmark expects small entries" - ); - if !zip_file.is_dir() { - let mut sink = [0_u8; 256]; - let bytes_read = - std::io::Read::read(&mut zip_file, &mut sink).expect("read zip entry payload for reader-only benchmark"); - total_bytes += bytes_read; - } - } - black_box(total_bytes); - }); - }); - - group.bench_function("file_write_only_stored_small", |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let extract_path = temp.path().join("extract"); - std::fs::create_dir_all(&extract_path).expect("create extract dir for file-write-only benchmark"); - let mut total_bytes = 0usize; - for index in 0..32 { - let path = extract_path.join(format!("file-{index}.txt")); - std::fs::write(&path, [b'c'; 256]).expect("write small file for file-write-only benchmark"); - total_bytes += 256; - } - black_box(total_bytes); - }); - }); - - group.finish(); -} - -fn build_object_archive_files( - metadata_count: usize, - metadata_size: usize, - payload_count: usize, - payload_size: usize, -) -> Vec<(String, Vec)> { - let mut files = Vec::with_capacity(metadata_count * 2 + payload_count); - - for index in 0..metadata_count { - let key_prefix = format!( - "bucket-a/shard-{}/tenant-{}/dataset-{}/object-{index:04}", - index % 8, - index % 16, - index % 32 - ); - files.push(( - format!("{key_prefix}/meta.json"), - format!( - "{{\"key\":\"object-{index:04}\",\"etag\":\"{:032x}\",\"size\":{},\"content_type\":\"application/octet-stream\"}}", - index, - payload_size - ) - .into_bytes(), - )); - files.push((format!("{key_prefix}/tags.txt"), vec![b'm'; metadata_size])); - } - - for index in 0..payload_count { - let payload_prefix = format!( - "bucket-a/shard-{}/tenant-{}/dataset-{}/object-{index:04}", - index % 8, - index % 16, - index % 32 - ); - files.push((format!("{payload_prefix}/part-00000.bin"), vec![b'p'; payload_size])); - } - - files -} - -fn bench_zip_object_archive_extract(c: &mut Criterion) { - let runtime = build_runtime(); - let mut group = c.benchmark_group("zip_object_archive_extract"); - - for (name, compression_level, metadata_count, metadata_size, payload_count, payload_size) in [ - ( - "stored_metadata_heavy_384m_24p", - CompressionLevel::Fastest, - 384usize, - 192usize, - 24usize, - 32 * 1024usize, - ), - ( - "deflated_mixed_192m_32p", - CompressionLevel::Best, - 192usize, - 256usize, - 32usize, - 64 * 1024usize, - ), - ] { - let files = build_object_archive_files(metadata_count, metadata_size, payload_count, payload_size); - let total_bytes = files.iter().map(|(_, payload)| payload.len() as u64).sum::(); - let payload = { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("object-archive.zip"); - runtime.block_on(async { - create_zip_with_options( - &zip_path, - files.clone(), - ZipWriteOptions { - compression_level, - create_directory_entries: true, - }, - ) - .await - .expect("create object archive benchmark payload"); - }); - std::fs::read(&zip_path).expect("read object archive benchmark payload") - }; - - group.throughput(Throughput::Bytes(total_bytes)); - group.bench_function(BenchmarkId::new("extract_full", name), |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - std::fs::write(&zip_path, &payload).expect("write object archive benchmark payload"); - runtime.block_on(async { - let entries = extract_zip_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("extract object archive benchmark payload"); - black_box(entries.len()); - }); - }); - }); - - group.bench_function(BenchmarkId::new("extract_summary_only", name), |b| { - b.iter(|| { - let temp = tempdir().expect("create benchmark tempdir"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - std::fs::write(&zip_path, &payload).expect("write object archive benchmark payload"); - runtime.block_on(async { - let summary = extract_zip_to_path_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("extract object archive benchmark summary"); - black_box(summary.file_count); - }); - }); - }); - } - - group.finish(); -} - -criterion_group!( - benches, - bench_tar_family_extract, - bench_zip_helper_round_trip, - bench_zip_helper_hotspot_breakdown, - bench_zip_object_archive_extract -); -criterion_main!(benches); diff --git a/crates/zip/src/lib.rs b/crates/zip/src/lib.rs index d61968fbb..82aa8867a 100644 --- a/crates/zip/src/lib.rs +++ b/crates/zip/src/lib.rs @@ -13,21 +13,8 @@ // limitations under the License. use async_compression::tokio::bufread::{BzDecoder, GzipDecoder, XzDecoder, ZlibDecoder, ZstdDecoder}; -use async_compression::tokio::write::{BzEncoder, GzipEncoder, XzEncoder, ZlibEncoder, ZstdEncoder}; -use std::collections::HashSet; -use std::future::Future; -use std::io::{Read, Write}; -use std::path::{Component, Path, PathBuf}; -use std::pin::Pin; -use std::sync::{Arc, Mutex}; -use std::task::{Context, Poll}; use thiserror::Error; -use tokio::fs::File; -use tokio::io::{self, AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt, BufReader, BufWriter}; -use tokio::task::spawn_blocking; -use tokio_stream::StreamExt; -use tokio_tar::Archive; -use zip::{CompressionMethod, ZipArchive, ZipWriter, write::SimpleFileOptions}; +use tokio::io::{AsyncRead, BufReader}; pub type Result = std::result::Result; @@ -38,51 +25,6 @@ pub enum ZipError { format: CompressionFormat, operation: &'static str, }, - #[error("invalid compression level {0}: value exceeds i32::MAX")] - InvalidCompressionLevel(u32), - #[error("unsafe archive entry path: {0}")] - UnsafeEntryPath(String), - #[error("archive entry path length {length} exceeds limit {limit}: {path}")] - EntryPathTooLong { path: String, length: usize, limit: usize }, - #[error("archive entry count {count} exceeds limit {limit}")] - EntryCountLimitExceeded { count: usize, limit: usize }, - #[error("archive entry '{path}' size {size} exceeds limit {limit}")] - EntrySizeLimitExceeded { path: String, size: u64, limit: u64 }, - #[error("archive total unpacked size {size} exceeds limit {limit}")] - TotalUnpackedSizeLimitExceeded { size: u64, limit: u64 }, - #[error(transparent)] - Io(#[from] io::Error), - #[error(transparent)] - Zip(#[from] zip::result::ZipError), - #[error(transparent)] - Join(#[from] tokio::task::JoinError), -} - -#[derive(Debug, PartialEq, Eq, Clone, Copy)] -pub enum CompressionCodec { - Gzip, - Bzip2, - Xz, - Zlib, - Zstd, -} - -#[derive(Debug, PartialEq, Eq, Clone, Copy)] -pub enum ArchiveKind { - Tar, - Zip, -} - -#[derive(Debug, PartialEq, Eq, Clone, Copy)] -pub enum ArchiveFormat { - Tar, - TarGzip, - TarBzip2, - TarXz, - TarZlib, - TarZstd, - Zip, - Unknown, } #[derive(Debug, PartialEq, Eq, Clone, Copy)] @@ -97,35 +39,9 @@ pub enum CompressionFormat { Unknown, } -#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)] -pub enum CompressionLevel { - Fastest, - Best, - #[default] - Default, - Level(u32), -} - -#[derive(Debug, Clone, PartialEq, Eq)] -pub struct ZipEntry { - pub name: String, - pub size: u64, - pub compressed_size: u64, - pub is_dir: bool, - pub compression_method: String, - pub archive_kind: ArchiveKind, - pub format: ArchiveFormat, - pub unix_mode: Option, -} - -#[derive(Debug, Clone, PartialEq, Eq, Default)] -pub struct ZipExtractSummary { - pub entry_count: usize, - pub directory_count: usize, - pub file_count: usize, - pub total_unpacked_size: u64, -} - +/// Archive guardrails. The values are carried here so every archive caller +/// shares one default policy; enforcement belongs to the caller, which maps a +/// breach onto its own protocol error. #[derive(Debug, Clone, Copy, PartialEq, Eq)] pub struct ArchiveLimits { pub max_entries: usize, @@ -147,103 +63,23 @@ impl Default for ArchiveLimits { } } -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub struct ZipWriteOptions { - pub compression_level: CompressionLevel, - pub create_directory_entries: bool, -} - -impl Default for ZipWriteOptions { - fn default() -> Self { - Self { - compression_level: CompressionLevel::Default, - create_directory_entries: false, - } - } -} - -const SMALL_ZIP_EXTRACT_FAST_PATH_LIMIT: u64 = 8 * 1024; - -#[derive(Clone, Default)] -struct SharedBuffer { - inner: Arc>>, -} - -impl SharedBuffer { - fn into_vec(self) -> Vec { - self.inner.lock().expect("shared in-memory writer lock poisoned").clone() - } -} - -impl AsyncWrite for SharedBuffer { - fn poll_write(self: Pin<&mut Self>, _cx: &mut Context<'_>, buf: &[u8]) -> Poll> { - let mut inner = self - .inner - .lock() - .map_err(|_| io::Error::other("shared in-memory writer lock poisoned"))?; - inner.extend_from_slice(buf); - Poll::Ready(Ok(buf.len())) - } - - fn poll_flush(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll> { - Poll::Ready(Ok(())) - } - - fn poll_shutdown(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll> { - Poll::Ready(Ok(())) - } -} - impl CompressionFormat { + /// Map an archive extension onto the stream codec needed to read it. + /// Tar-family suffixes (`tgz`, `tbz2`, `txz`, `tzst`, ...) resolve to their + /// codec because the tar container itself is read from the decoded stream. pub fn from_extension(ext: &str) -> Self { - Self::from_archive_format(ArchiveFormat::from_extension(ext)) - } - - pub fn from_archive_format(format: ArchiveFormat) -> Self { - match format { - ArchiveFormat::TarGzip => CompressionFormat::Gzip, - ArchiveFormat::TarBzip2 => CompressionFormat::Bzip2, - ArchiveFormat::TarXz => CompressionFormat::Xz, - ArchiveFormat::TarZlib => CompressionFormat::Zlib, - ArchiveFormat::TarZstd => CompressionFormat::Zstd, - ArchiveFormat::Tar => CompressionFormat::Tar, - ArchiveFormat::Zip => CompressionFormat::Zip, - ArchiveFormat::Unknown => CompressionFormat::Unknown, + match ext.to_ascii_lowercase().as_str() { + "gz" | "gzip" | "tgz" => CompressionFormat::Gzip, + "bz2" | "bzip2" | "tbz" | "tbz2" => CompressionFormat::Bzip2, + "xz" | "txz" => CompressionFormat::Xz, + "zlib" | "zz" => CompressionFormat::Zlib, + "zst" | "zstd" | "tzst" => CompressionFormat::Zstd, + "tar" => CompressionFormat::Tar, + "zip" => CompressionFormat::Zip, + _ => CompressionFormat::Unknown, } } - pub fn archive_format_from_path>(path: P) -> ArchiveFormat { - ArchiveFormat::from_path(path) - } - - pub fn archive_kind(&self) -> Option { - match self { - CompressionFormat::Tar => Some(ArchiveKind::Tar), - CompressionFormat::Zip => Some(ArchiveKind::Zip), - CompressionFormat::Gzip - | CompressionFormat::Bzip2 - | CompressionFormat::Xz - | CompressionFormat::Zlib - | CompressionFormat::Zstd - | CompressionFormat::Unknown => None, - } - } - - pub fn compression_codec(&self) -> Option { - match self { - CompressionFormat::Gzip => Some(CompressionCodec::Gzip), - CompressionFormat::Bzip2 => Some(CompressionCodec::Bzip2), - CompressionFormat::Xz => Some(CompressionCodec::Xz), - CompressionFormat::Zlib => Some(CompressionCodec::Zlib), - CompressionFormat::Zstd => Some(CompressionCodec::Zstd), - CompressionFormat::Tar | CompressionFormat::Zip | CompressionFormat::Unknown => None, - } - } - - pub fn from_path>(path: P) -> Self { - Self::from_archive_format(ArchiveFormat::from_path(path)) - } - pub fn extension(&self) -> &'static str { match self { CompressionFormat::Gzip => "gz", @@ -257,10 +93,6 @@ impl CompressionFormat { } } - pub fn is_supported(&self) -> bool { - !matches!(self, CompressionFormat::Unknown) - } - pub fn get_decoder(&self, input: R) -> Result> where R: AsyncRead + Send + Unpin + 'static, @@ -290,625 +122,14 @@ impl CompressionFormat { Ok(decoder) } - - fn convert_level(level: CompressionLevel) -> Result { - match level { - CompressionLevel::Fastest => Ok(async_compression::Level::Fastest), - CompressionLevel::Best => Ok(async_compression::Level::Best), - CompressionLevel::Default => Ok(async_compression::Level::Default), - CompressionLevel::Level(n) => { - let level = i32::try_from(n).map_err(|_| ZipError::InvalidCompressionLevel(n))?; - Ok(async_compression::Level::Precise(level)) - } - } - } - - pub fn get_encoder(&self, output: W, level: CompressionLevel) -> Result> - where - W: AsyncWrite + Send + Unpin + 'static, - { - let writer = BufWriter::new(output); - - let encoder: Box = match self { - CompressionFormat::Gzip => Box::new(GzipEncoder::with_quality(writer, Self::convert_level(level)?)), - CompressionFormat::Bzip2 => Box::new(BzEncoder::with_quality(writer, Self::convert_level(level)?)), - CompressionFormat::Zlib => Box::new(ZlibEncoder::with_quality(writer, Self::convert_level(level)?)), - CompressionFormat::Xz => Box::new(XzEncoder::with_quality(writer, Self::convert_level(level)?)), - CompressionFormat::Zstd => Box::new(ZstdEncoder::with_quality(writer, Self::convert_level(level)?)), - CompressionFormat::Tar => Box::new(writer), - CompressionFormat::Zip => { - return Err(ZipError::UnsupportedFormat { - format: *self, - operation: "stream encoding", - }); - } - CompressionFormat::Unknown => { - return Err(ZipError::UnsupportedFormat { - format: *self, - operation: "encoding", - }); - } - }; - - Ok(encoder) - } -} - -impl ArchiveFormat { - pub fn from_extension(ext: &str) -> Self { - match ext.to_ascii_lowercase().as_str() { - "gz" | "gzip" | "tgz" => ArchiveFormat::TarGzip, - "bz2" | "bzip2" | "tbz" | "tbz2" => ArchiveFormat::TarBzip2, - "xz" | "txz" => ArchiveFormat::TarXz, - "zlib" | "zz" => ArchiveFormat::TarZlib, - "zst" | "zstd" | "tzst" => ArchiveFormat::TarZstd, - "tar" => ArchiveFormat::Tar, - "zip" => ArchiveFormat::Zip, - _ => ArchiveFormat::Unknown, - } - } - - pub fn from_path>(path: P) -> Self { - let path = path.as_ref(); - let lower_name = path.file_name().and_then(|name| name.to_str()).map(str::to_ascii_lowercase); - - if let Some(name) = lower_name { - if name.ends_with(".tar.gz") || name.ends_with(".tgz") { - return ArchiveFormat::TarGzip; - } - if name.ends_with(".tar.bz2") || name.ends_with(".tbz") || name.ends_with(".tbz2") { - return ArchiveFormat::TarBzip2; - } - if name.ends_with(".tar.xz") || name.ends_with(".txz") { - return ArchiveFormat::TarXz; - } - if name.ends_with(".tar.zst") || name.ends_with(".tzst") { - return ArchiveFormat::TarZstd; - } - if name.ends_with(".tar.zlib") { - return ArchiveFormat::TarZlib; - } - } - - path.extension() - .and_then(|s| s.to_str()) - .map(Self::from_extension) - .unwrap_or(ArchiveFormat::Unknown) - } - - pub fn archive_kind(&self) -> Option { - match self { - ArchiveFormat::Tar - | ArchiveFormat::TarGzip - | ArchiveFormat::TarBzip2 - | ArchiveFormat::TarXz - | ArchiveFormat::TarZlib - | ArchiveFormat::TarZstd => Some(ArchiveKind::Tar), - ArchiveFormat::Zip => Some(ArchiveKind::Zip), - ArchiveFormat::Unknown => None, - } - } - - pub fn compression_codec(&self) -> Option { - match self { - ArchiveFormat::TarGzip => Some(CompressionCodec::Gzip), - ArchiveFormat::TarBzip2 => Some(CompressionCodec::Bzip2), - ArchiveFormat::TarXz => Some(CompressionCodec::Xz), - ArchiveFormat::TarZlib => Some(CompressionCodec::Zlib), - ArchiveFormat::TarZstd => Some(CompressionCodec::Zstd), - ArchiveFormat::Tar | ArchiveFormat::Zip | ArchiveFormat::Unknown => None, - } - } - - pub fn extension(&self) -> &'static str { - match self { - ArchiveFormat::Tar => "tar", - ArchiveFormat::TarGzip => "tar.gz", - ArchiveFormat::TarBzip2 => "tar.bz2", - ArchiveFormat::TarXz => "tar.xz", - ArchiveFormat::TarZlib => "tar.zlib", - ArchiveFormat::TarZstd => "tar.zst", - ArchiveFormat::Zip => "zip", - ArchiveFormat::Unknown => "", - } - } -} - -/// Read entries from a tar-family archive stream. -/// -/// Supported formats are: -/// - `CompressionFormat::Tar` -/// - `CompressionFormat::Gzip` -/// - `CompressionFormat::Bzip2` -/// - `CompressionFormat::Xz` -/// - `CompressionFormat::Zlib` -/// - `CompressionFormat::Zstd` -/// -/// `CompressionFormat::Zip` is intentionally not supported here because ZIP -/// requires central-directory semantics and is handled through file-based -/// helper APIs. -pub async fn read_archive_entries(input: R, format: CompressionFormat, callback: F) -> Result<()> -where - R: AsyncRead + Send + Unpin + 'static, - F: FnMut(tokio_tar::Entry>>) -> Fut + Send + 'static, - Fut: Future> + Send, -{ - read_archive_entries_with_limits(input, format, ArchiveLimits::default(), callback).await -} - -pub async fn read_archive_entries_with_limits( - input: R, - format: CompressionFormat, - limits: ArchiveLimits, - mut callback: F, -) -> Result<()> -where - R: AsyncRead + Send + Unpin + 'static, - F: FnMut(tokio_tar::Entry>>) -> Fut + Send + 'static, - Fut: Future> + Send, -{ - let decoder = format.get_decoder(input)?; - let mut ar = Archive::new(decoder); - let mut entries = ar.entries()?; - let mut entry_count = 0_usize; - let mut total_unpacked_size = 0_u64; - - while let Some(entry) = entries.next().await { - let entry = entry?; - entry_count += 1; - validate_archive_entry_count(entry_count, limits)?; - - let entry_path = entry.path()?.to_string_lossy().into_owned(); - validate_archive_entry_name(&entry_path, limits)?; - - let entry_size = entry.header().size()?; - validate_archive_entry_size(&entry_path, entry_size, limits)?; - total_unpacked_size = total_unpacked_size.saturating_add(entry_size); - validate_archive_total_size(total_unpacked_size, limits)?; - - callback(entry).await?; - } - - Ok(()) -} - -/// Backward-compatible wrapper for archive entry iteration. -pub async fn decompress(input: R, format: CompressionFormat, callback: F) -> Result<()> -where - R: AsyncRead + Send + Unpin + 'static, - F: FnMut(tokio_tar::Entry>>) -> Fut + Send + 'static, - Fut: Future> + Send, -{ - read_archive_entries(input, format, callback).await -} - -/// Explicit tar-family alias for callers that want a clearer name than -/// `decompress()`. -pub async fn extract_tar_entries(input: R, format: CompressionFormat, callback: F) -> Result<()> -where - R: AsyncRead + Send + Unpin + 'static, - F: FnMut(tokio_tar::Entry>>) -> Fut + Send + 'static, - Fut: Future> + Send, -{ - read_archive_entries(input, format, callback).await -} - -fn normalize_zip_entry_name(name: &str) -> Result { - let path = Path::new(name); - let mut normalized = PathBuf::new(); - - for component in path.components() { - match component { - Component::Normal(part) => normalized.push(part), - Component::CurDir => {} - Component::ParentDir | Component::RootDir | Component::Prefix(_) => { - return Err(ZipError::UnsafeEntryPath(name.to_string())); - } - } - } - - let normalized = normalized.to_string_lossy().replace('\\', "/"); - if normalized.is_empty() { - return Err(ZipError::UnsafeEntryPath(name.to_string())); - } - - Ok(normalized) -} - -fn validate_archive_entry_name(name: &str, limits: ArchiveLimits) -> Result<()> { - if !limits.validate_entry_paths { - return Ok(()); - } - - let normalized = normalize_zip_entry_name(name)?; - let length = normalized.len(); - if length > limits.max_path_length { - return Err(ZipError::EntryPathTooLong { - path: normalized, - length, - limit: limits.max_path_length, - }); - } - - Ok(()) -} - -fn validate_archive_entry_size(path: &str, size: u64, limits: ArchiveLimits) -> Result<()> { - if size > limits.max_entry_size { - return Err(ZipError::EntrySizeLimitExceeded { - path: path.to_string(), - size, - limit: limits.max_entry_size, - }); - } - - Ok(()) -} - -fn validate_archive_entry_count(count: usize, limits: ArchiveLimits) -> Result<()> { - if count > limits.max_entries { - return Err(ZipError::EntryCountLimitExceeded { - count, - limit: limits.max_entries, - }); - } - - Ok(()) -} - -fn validate_archive_total_size(total_size: u64, limits: ArchiveLimits) -> Result<()> { - if total_size > limits.max_total_unpacked_size { - return Err(ZipError::TotalUnpackedSizeLimitExceeded { - size: total_size, - limit: limits.max_total_unpacked_size, - }); - } - - Ok(()) -} - -fn zip_method_for_level(level: CompressionLevel) -> CompressionMethod { - match level { - CompressionLevel::Fastest => CompressionMethod::Stored, - CompressionLevel::Best | CompressionLevel::Default | CompressionLevel::Level(_) => CompressionMethod::Deflated, - } -} - -fn parent_directories_for(path: &str) -> Vec { - let path = Path::new(path); - let mut current = PathBuf::new(); - let mut directories = Vec::new(); - - if let Some(parent) = path.parent() { - for component in parent.components() { - if let Component::Normal(part) = component { - current.push(part); - directories.push(format!("{}/", current.to_string_lossy().replace('\\', "/"))); - } - } - } - - directories -} - -fn ensure_directory(path: &Path, created_directories: &mut HashSet) -> Result<()> { - let path = path.to_path_buf(); - if created_directories.insert(path.clone()) { - std::fs::create_dir_all(&path)?; - } - - Ok(()) -} - -fn write_small_zip_entry(reader: &mut R, output_path: &Path, size: u64) -> Result<()> { - let size = usize::try_from(size).map_err(|_| io::Error::other("small zip entry size overflow"))?; - let mut buffer = [0_u8; SMALL_ZIP_EXTRACT_FAST_PATH_LIMIT as usize]; - reader.read_exact(&mut buffer[..size])?; - // `read_exact` stops as soon as the declared bytes are read and never performs the - // terminal zero-length read that the zip crate's `Crc32Reader` uses to validate the - // entry checksum. Force one extra read to EOF so a corrupted small entry is rejected - // here, matching the large-entry `io::copy` path which already reads through EOF. - let mut trailing = [0_u8; 1]; - if reader.read(&mut trailing)? != 0 { - return Err(io::Error::other("small zip entry produced more data than its declared size").into()); - } - std::fs::write(output_path, &buffer[..size])?; - Ok(()) -} - -pub async fn extract_zip_simple, Q: AsRef>(zip_path: P, extract_to: Q) -> Result> { - extract_zip_with_limits(zip_path, extract_to, ArchiveLimits::default()).await -} - -pub async fn extract_zip_to_path_with_limits, Q: AsRef>( - zip_path: P, - extract_to: Q, - limits: ArchiveLimits, -) -> Result { - let zip_path = zip_path.as_ref().to_path_buf(); - let extract_to = extract_to.as_ref().to_path_buf(); - - spawn_blocking(move || extract_zip_impl(zip_path, extract_to, limits, false).map(|(_, summary)| summary)).await? -} - -pub async fn extract_zip_with_limits, Q: AsRef>( - zip_path: P, - extract_to: Q, - limits: ArchiveLimits, -) -> Result> { - let zip_path = zip_path.as_ref().to_path_buf(); - let extract_to = extract_to.as_ref().to_path_buf(); - - spawn_blocking(move || extract_zip_impl(zip_path, extract_to, limits, true).map(|(entries, _)| entries.unwrap_or_default())) - .await? -} - -fn extract_zip_impl( - zip_path: PathBuf, - extract_to: PathBuf, - limits: ArchiveLimits, - collect_entries: bool, -) -> Result<(Option>, ZipExtractSummary)> { - let file = std::fs::File::open(&zip_path)?; - let mut archive = ZipArchive::new(file)?; - std::fs::create_dir_all(&extract_to)?; - let mut created_directories = HashSet::from([extract_to.clone()]); - - let mut entries = collect_entries.then(|| Vec::with_capacity(archive.len())); - let mut summary = ZipExtractSummary::default(); - for index in 0..archive.len() { - validate_archive_entry_count(index + 1, limits)?; - let mut zip_file = archive.by_index(index)?; - let enclosed_name = zip_file - .enclosed_name() - .ok_or_else(|| ZipError::UnsafeEntryPath(zip_file.name().to_string()))?; - let entry_name = enclosed_name.to_string_lossy().replace('\\', "/"); - let is_dir = zip_file.is_dir(); - let size = zip_file.size(); - validate_archive_entry_name(&entry_name, limits)?; - validate_archive_entry_size(&entry_name, size, limits)?; - summary.total_unpacked_size = summary.total_unpacked_size.saturating_add(size); - validate_archive_total_size(summary.total_unpacked_size, limits)?; - let output_path = extract_to.join(&enclosed_name); - - if is_dir { - ensure_directory(&output_path, &mut created_directories)?; - summary.directory_count += 1; - } else { - if let Some(parent) = output_path.parent() { - ensure_directory(parent, &mut created_directories)?; - } - if size <= SMALL_ZIP_EXTRACT_FAST_PATH_LIMIT { - write_small_zip_entry(&mut zip_file, &output_path, size)?; - } else { - let mut output = std::fs::File::create(&output_path)?; - std::io::copy(&mut zip_file, &mut output)?; - } - summary.file_count += 1; - } - summary.entry_count += 1; - - if let Some(ref mut entries) = entries { - entries.push(ZipEntry { - name: entry_name, - size, - compressed_size: zip_file.compressed_size(), - is_dir, - compression_method: format!("{:?}", zip_file.compression()), - archive_kind: ArchiveKind::Zip, - format: ArchiveFormat::Zip, - unix_mode: zip_file.unix_mode(), - }); - } - } - - Ok((entries, summary)) -} - -pub async fn create_zip_simple>( - zip_path: P, - files: Vec<(String, Vec)>, - compression_level: CompressionLevel, -) -> Result<()> { - create_zip_with_options( - zip_path, - files, - ZipWriteOptions { - compression_level, - ..ZipWriteOptions::default() - }, - ) - .await -} - -pub async fn create_zip_with_options>( - zip_path: P, - files: Vec<(String, Vec)>, - options: ZipWriteOptions, -) -> Result<()> { - let zip_path = zip_path.as_ref().to_path_buf(); - - spawn_blocking(move || -> Result<()> { - if let Some(parent) = zip_path.parent() { - std::fs::create_dir_all(parent)?; - } - - let file = std::fs::File::create(&zip_path)?; - let mut writer = ZipWriter::new(file); - let file_options = SimpleFileOptions::default().compression_method(zip_method_for_level(options.compression_level)); - let explicit_directories = files - .iter() - .filter(|(name, _)| name.ends_with('/')) - .map(|(name, _)| normalize_zip_entry_name(name)) - .collect::>>()?; - let mut written_directories = HashSet::new(); - - for (name, contents) in files { - let entry_name = normalize_zip_entry_name(&name)?; - if name.ends_with('/') { - if written_directories.insert(entry_name.clone()) { - writer.add_directory(entry_name, file_options)?; - } - } else { - if options.create_directory_entries { - for directory in parent_directories_for(&entry_name) { - if !explicit_directories.contains(&directory) && written_directories.insert(directory.clone()) { - writer.add_directory(directory, file_options)?; - } - } - } - writer.start_file(entry_name, file_options)?; - writer.write_all(&contents)?; - } - } - - writer.finish()?; - Ok(()) - }) - .await? -} - -pub struct Compressor { - format: CompressionFormat, - level: CompressionLevel, -} - -impl Compressor { - pub fn new(format: CompressionFormat) -> Self { - Self { - format, - level: CompressionLevel::Default, - } - } - - pub fn with_level(mut self, level: CompressionLevel) -> Self { - self.level = level; - self - } - - pub async fn compress(&self, input: &[u8]) -> Result> { - let sink = SharedBuffer::default(); - let mut encoder = self.format.get_encoder(sink.clone(), self.level)?; - let mut reader = input; - - io::copy(&mut reader, &mut encoder).await?; - encoder.shutdown().await?; - drop(encoder); - - Ok(sink.into_vec()) - } - - pub async fn decompress(&self, input: Vec) -> Result> { - let mut output = Vec::new(); - let cursor = std::io::Cursor::new(input); - let mut decoder = self.format.get_decoder(cursor)?; - - decoder.read_to_end(&mut output).await?; - Ok(output) - } -} - -pub struct Decompressor { - format: CompressionFormat, -} - -impl Decompressor { - pub fn new(format: CompressionFormat) -> Self { - Self { format } - } - - pub fn auto_detect>(path: P) -> Self { - Self { - format: CompressionFormat::from_path(path), - } - } - - pub async fn decompress_file>(&self, input_path: P, output_path: P) -> Result<()> { - let input_file = File::open(&input_path).await?; - let output_file = File::create(&output_path).await?; - - let mut decoder = self.format.get_decoder(input_file)?; - let mut writer = BufWriter::new(output_file); - - io::copy(&mut decoder, &mut writer).await?; - writer.shutdown().await?; - - Ok(()) - } } #[cfg(test)] mod tests { use super::*; + use async_compression::tokio::write::GzipEncoder; use std::mem::size_of; - use tempfile::tempdir; - use tokio::fs; - use tokio::io::AsyncReadExt; - use tokio_tar::{Builder, Header}; - use zip::write::FileOptions; - - async fn build_tar_bytes(files: &[(&str, &[u8])]) -> io::Result> { - let sink = SharedBuffer::default(); - let handle = sink.clone(); - let mut builder = Builder::new(sink); - - for (path, content) in files { - let mut header = Header::new_gnu(); - header.set_size(content.len() as u64); - header.set_mode(0o644); - header.set_cksum(); - builder.append_data(&mut header, *path, &content[..]).await?; - } - - builder.finish().await?; - Ok(handle.into_vec()) - } - - async fn build_compressed_tar_bytes(format: CompressionFormat, files: &[(&str, &[u8])]) -> Result> { - let tar_bytes = build_tar_bytes(files).await?; - Compressor::new(format).compress(&tar_bytes).await - } - - async fn build_zip_file_with_entries(path: &Path, files: &[(&str, &[u8])]) -> Result<()> { - let path = path.to_path_buf(); - let files = files - .iter() - .map(|(name, content)| ((*name).to_string(), content.to_vec())) - .collect::>(); - spawn_blocking(move || -> Result<()> { - let file = std::fs::File::create(path)?; - let mut writer = ZipWriter::new(file); - let options: FileOptions<'_, ()> = FileOptions::default().compression_method(CompressionMethod::Stored); - for (name, content) in files { - writer.start_file(name, options)?; - writer.write_all(&content)?; - } - writer.finish()?; - Ok(()) - }) - .await??; - Ok(()) - } - - async fn collect_archive_entries(payload: Vec, format: CompressionFormat) -> Result)>> { - let seen = Arc::new(Mutex::new(Vec::<(String, Vec)>::new())); - let seen_ref = Arc::clone(&seen); - let cursor = std::io::Cursor::new(payload); - - read_archive_entries(cursor, format, move |mut entry| { - let seen_ref = Arc::clone(&seen_ref); - async move { - let path = entry.path()?.to_string_lossy().into_owned(); - let mut content = Vec::new(); - entry.read_to_end(&mut content).await?; - seen_ref.lock().expect("seen collection lock poisoned").push((path, content)); - Ok(()) - } - }) - .await?; - - Ok(seen.lock().expect("seen collection lock poisoned").clone()) - } + use tokio::io::{AsyncReadExt, AsyncWriteExt}; #[test] fn test_compression_format_from_extension() { @@ -918,869 +139,51 @@ mod tests { assert_eq!(CompressionFormat::from_extension("txt"), CompressionFormat::Unknown); } - #[test] - fn test_archive_format_from_extension() { - assert_eq!(ArchiveFormat::from_extension("gz"), ArchiveFormat::TarGzip); - assert_eq!(ArchiveFormat::from_extension("tbz2"), ArchiveFormat::TarBzip2); - assert_eq!(ArchiveFormat::from_extension("txz"), ArchiveFormat::TarXz); - assert_eq!(ArchiveFormat::from_extension("zip"), ArchiveFormat::Zip); - assert_eq!(ArchiveFormat::from_extension("txt"), ArchiveFormat::Unknown); - } - - #[test] - fn test_compression_format_from_path_handles_compound_suffixes() { - assert_eq!(CompressionFormat::from_path("archive.tar.gz"), CompressionFormat::Gzip); - assert_eq!(CompressionFormat::from_path("archive.tgz"), CompressionFormat::Gzip); - assert_eq!(CompressionFormat::from_path("archive.tar.bz2"), CompressionFormat::Bzip2); - assert_eq!(CompressionFormat::from_path("archive.zip"), CompressionFormat::Zip); - assert_eq!(CompressionFormat::from_path("archive"), CompressionFormat::Unknown); - } - - #[test] - fn test_archive_format_from_path_handles_compound_suffixes() { - assert_eq!(ArchiveFormat::from_path("archive.tar.gz"), ArchiveFormat::TarGzip); - assert_eq!(ArchiveFormat::from_path("archive.tar.bz2"), ArchiveFormat::TarBzip2); - assert_eq!(ArchiveFormat::from_path("archive.tar.xz"), ArchiveFormat::TarXz); - assert_eq!(ArchiveFormat::from_path("archive.tar.zst"), ArchiveFormat::TarZstd); - assert_eq!(ArchiveFormat::from_path("archive.zip"), ArchiveFormat::Zip); - assert_eq!(ArchiveFormat::from_path("archive"), ArchiveFormat::Unknown); - } - - #[test] - fn test_archive_format_and_legacy_compression_format_are_compatible() { - assert_eq!(CompressionFormat::from_archive_format(ArchiveFormat::TarGzip), CompressionFormat::Gzip); - assert_eq!(CompressionFormat::from_archive_format(ArchiveFormat::Tar), CompressionFormat::Tar); - assert_eq!(CompressionFormat::from_archive_format(ArchiveFormat::Zip), CompressionFormat::Zip); - } - - #[test] - fn test_archive_format_exposes_archive_kind_and_codec() { - assert_eq!(ArchiveFormat::TarGzip.archive_kind(), Some(ArchiveKind::Tar)); - assert_eq!(ArchiveFormat::TarGzip.compression_codec(), Some(CompressionCodec::Gzip)); - assert_eq!(ArchiveFormat::Tar.archive_kind(), Some(ArchiveKind::Tar)); - assert_eq!(ArchiveFormat::Tar.compression_codec(), None); - assert_eq!(ArchiveFormat::Zip.archive_kind(), Some(ArchiveKind::Zip)); - assert_eq!(ArchiveFormat::Zip.compression_codec(), None); - } - - #[test] - fn test_legacy_compression_format_exposes_kind_and_codec() { - assert_eq!(CompressionFormat::Gzip.archive_kind(), None); - assert_eq!(CompressionFormat::Gzip.compression_codec(), Some(CompressionCodec::Gzip)); - assert_eq!(CompressionFormat::Tar.archive_kind(), Some(ArchiveKind::Tar)); - assert_eq!(CompressionFormat::Tar.compression_codec(), None); - assert_eq!(CompressionFormat::Zip.archive_kind(), Some(ArchiveKind::Zip)); - } - #[test] fn test_compression_format_size_is_small() { assert!(size_of::() <= 8); assert!(size_of::>() <= 16); } - #[test] - fn test_convert_level_rejects_overflow() { - let err = match CompressionFormat::Gzip.get_encoder(SharedBuffer::default(), CompressionLevel::Level(u32::MAX)) { - Ok(_) => panic!("overflow level should return an error"), - Err(err) => err, - }; - assert!(matches!(err, ZipError::InvalidCompressionLevel(u32::MAX))); - } + #[tokio::test] + async fn test_get_decoder_round_trips_gzip_stream() { + let mut encoder = GzipEncoder::new(Vec::new()); + encoder.write_all(b"payload").await.expect("gzip encode should succeed"); + encoder.shutdown().await.expect("gzip encoder shutdown should succeed"); - #[test] - fn test_validate_archive_entry_name_rejects_absolute_path() { - let err = validate_archive_entry_name("/absolute.txt", ArchiveLimits::default()) - .expect_err("absolute path should fail validation"); - assert!(matches!(err, ZipError::UnsafeEntryPath(path) if path == "/absolute.txt")); + let mut decoder = CompressionFormat::Gzip + .get_decoder(std::io::Cursor::new(encoder.into_inner())) + .expect("gzip decoder should be created"); + let mut decoded = Vec::new(); + decoder.read_to_end(&mut decoded).await.expect("gzip decode should succeed"); + + assert_eq!(decoded, b"payload"); } #[tokio::test] - async fn test_compressor_round_trip_gzip() { - let input = b"hello rustfs zip ".repeat(64); - let compressor = Compressor::new(CompressionFormat::Gzip); - - let compressed = compressor.compress(&input).await.expect("gzip compress should succeed"); - assert!(!compressed.is_empty()); - assert_ne!(compressed, input); - - let decompressed = compressor - .decompress(compressed) - .await - .expect("gzip decompress should succeed"); - assert_eq!(decompressed, input); - } - - #[tokio::test] - async fn test_compressor_round_trip_zstd() { - let input = b"zstd payload ".repeat(128); - let compressor = Compressor::new(CompressionFormat::Zstd).with_level(CompressionLevel::Best); - - let compressed = compressor.compress(&input).await.expect("zstd compress should succeed"); - let decompressed = compressor - .decompress(compressed) - .await - .expect("zstd decompress should succeed"); - assert_eq!(decompressed, input); - } - - #[tokio::test] - async fn test_zip_stream_encoder_is_rejected() { - let err = CompressionFormat::Zip - .get_encoder(SharedBuffer::default(), CompressionLevel::Default) + async fn test_get_decoder_rejects_zip_and_unknown_formats() { + let zip_err = CompressionFormat::Zip + .get_decoder(std::io::Cursor::new(Vec::::new())) .err() - .expect("zip stream encoder should be rejected"); + .expect("zip stream decoding should be rejected"); assert!(matches!( - err, - ZipError::UnsupportedFormat { - format: CompressionFormat::Zip, - operation: "stream encoding", - } - )); - } - - #[tokio::test] - async fn test_read_archive_entries_iterates_tar_gzip_entries() { - let gzip_bytes = - build_compressed_tar_bytes(CompressionFormat::Gzip, &[("nested/hello.txt", b"hello"), ("world.txt", b"world")]) - .await - .expect("tar.gz build should succeed"); - - let seen = collect_archive_entries(gzip_bytes, CompressionFormat::Gzip) - .await - .expect("tar.gz archive iteration should succeed"); - assert_eq!(seen.len(), 2); - assert_eq!(seen[0].0, "nested/hello.txt"); - assert_eq!(seen[0].1, b"hello"); - assert_eq!(seen[1].0, "world.txt"); - assert_eq!(seen[1].1, b"world"); - } - - #[tokio::test] - async fn test_read_archive_entries_iterates_tar_bzip2_entries() { - let payload = - build_compressed_tar_bytes(CompressionFormat::Bzip2, &[("nested/hello.txt", b"hello"), ("world.txt", b"world")]) - .await - .expect("tar.bz2 build should succeed"); - - let seen = collect_archive_entries(payload, CompressionFormat::Bzip2) - .await - .expect("tar.bz2 archive iteration should succeed"); - assert_eq!(seen.len(), 2); - assert_eq!(seen[0].0, "nested/hello.txt"); - assert_eq!(seen[1].0, "world.txt"); - } - - #[tokio::test] - async fn test_read_archive_entries_iterates_tar_xz_entries() { - let payload = - build_compressed_tar_bytes(CompressionFormat::Xz, &[("nested/hello.txt", b"hello"), ("world.txt", b"world")]) - .await - .expect("tar.xz build should succeed"); - - let seen = collect_archive_entries(payload, CompressionFormat::Xz) - .await - .expect("tar.xz archive iteration should succeed"); - assert_eq!(seen.len(), 2); - assert_eq!(seen[0].0, "nested/hello.txt"); - assert_eq!(seen[1].0, "world.txt"); - } - - #[tokio::test] - async fn test_read_archive_entries_iterates_tar_zstd_entries() { - let payload = - build_compressed_tar_bytes(CompressionFormat::Zstd, &[("nested/hello.txt", b"hello"), ("world.txt", b"world")]) - .await - .expect("tar.zst build should succeed"); - - let seen = collect_archive_entries(payload, CompressionFormat::Zstd) - .await - .expect("tar.zst archive iteration should succeed"); - assert_eq!(seen.len(), 2); - assert_eq!(seen[0].0, "nested/hello.txt"); - assert_eq!(seen[1].0, "world.txt"); - } - - #[tokio::test] - async fn test_extract_tar_entries_alias_matches_stream_behavior() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("hello.txt", b"hello")]) - .await - .expect("tar.gz build should succeed"); - let seen = Arc::new(Mutex::new(Vec::::new())); - let seen_ref = Arc::clone(&seen); - - extract_tar_entries(std::io::Cursor::new(payload), CompressionFormat::Gzip, move |entry| { - let seen_ref = Arc::clone(&seen_ref); - async move { - seen_ref - .lock() - .expect("seen collection lock poisoned") - .push(entry.path()?.to_string_lossy().into_owned()); - Ok(()) - } - }) - .await - .expect("extract_tar_entries alias should succeed"); - - assert_eq!(seen.lock().expect("seen collection lock poisoned").as_slice(), ["hello.txt"]); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_zip_streams() { - let err = read_archive_entries(std::io::Cursor::new(Vec::::new()), CompressionFormat::Zip, |_entry| async { Ok(()) }) - .await - .expect_err("zip stream should be rejected"); - - assert!(matches!( - err, + zip_err, ZipError::UnsupportedFormat { format: CompressionFormat::Zip, operation: "stream decoding", } )); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_corrupt_tar_gzip_stream() { - let err = read_archive_entries( - std::io::Cursor::new(b"not-a-valid-gzip-stream".to_vec()), - CompressionFormat::Gzip, - |_entry| async { Ok(()) }, - ) - .await - .expect_err("corrupt tar.gz stream should fail"); - - assert!(matches!(err, ZipError::Io(_))); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_truncated_tar_gzip_stream() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("hello.txt", b"hello world")]) - .await - .expect("tar.gz build should succeed"); - let truncated = payload[..payload.len() / 2].to_vec(); - - let err = read_archive_entries(std::io::Cursor::new(truncated), CompressionFormat::Gzip, |_entry| async { Ok(()) }) - .await - .expect_err("truncated tar.gz stream should fail"); - - assert!(matches!(err, ZipError::Io(_))); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_too_many_entries() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("one.txt", b"1"), ("two.txt", b"2")]) - .await - .expect("tar.gz build should succeed"); - - let err = read_archive_entries_with_limits( - std::io::Cursor::new(payload), - CompressionFormat::Gzip, - ArchiveLimits { - max_entries: 1, - ..ArchiveLimits::default() - }, - |_entry| async { Ok(()) }, - ) - .await - .expect_err("entry count limit should fail"); - - assert!(matches!(err, ZipError::EntryCountLimitExceeded { count: 2, limit: 1 })); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_oversized_entry() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("big.txt", b"hello world")]) - .await - .expect("tar.gz build should succeed"); - - let err = read_archive_entries_with_limits( - std::io::Cursor::new(payload), - CompressionFormat::Gzip, - ArchiveLimits { - max_entry_size: 4, - ..ArchiveLimits::default() - }, - |_entry| async { Ok(()) }, - ) - .await - .expect_err("entry size limit should fail"); + let unknown_err = CompressionFormat::Unknown + .get_decoder(std::io::Cursor::new(Vec::::new())) + .err() + .expect("unknown format decoding should be rejected"); assert!(matches!( - err, - ZipError::EntrySizeLimitExceeded { - path, - size: 11, - limit: 4, - } if path == "big.txt" + unknown_err, + ZipError::UnsupportedFormat { + format: CompressionFormat::Unknown, + operation: "decoding", + } )); } - - #[tokio::test] - async fn test_read_archive_entries_rejects_total_unpacked_size_limit() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("one.txt", b"12345"), ("two.txt", b"67890")]) - .await - .expect("tar.gz build should succeed"); - - let err = read_archive_entries_with_limits( - std::io::Cursor::new(payload), - CompressionFormat::Gzip, - ArchiveLimits { - max_total_unpacked_size: 9, - ..ArchiveLimits::default() - }, - |_entry| async { Ok(()) }, - ) - .await - .expect_err("total unpacked size limit should fail"); - - assert!(matches!(err, ZipError::TotalUnpackedSizeLimitExceeded { size: 10, limit: 9 })); - } - - #[tokio::test] - async fn test_read_archive_entries_rejects_entry_path_length_limit() { - let payload = build_compressed_tar_bytes(CompressionFormat::Gzip, &[("nested/hello.txt", b"hello")]) - .await - .expect("tar.gz build should succeed"); - - let err = read_archive_entries_with_limits( - std::io::Cursor::new(payload), - CompressionFormat::Gzip, - ArchiveLimits { - max_path_length: 5, - ..ArchiveLimits::default() - }, - |_entry| async { Ok(()) }, - ) - .await - .expect_err("path length limit should fail"); - - assert!(matches!( - err, - ZipError::EntryPathTooLong { path, limit: 5, .. } if path == "nested/hello.txt" - )); - } - - #[tokio::test] - async fn test_create_and_extract_zip_round_trip() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("archive.zip"); - let extract_path = temp.path().join("extract"); - - create_zip_simple( - &zip_path, - vec![ - ("nested/hello.txt".to_string(), b"hello".to_vec()), - ("world.txt".to_string(), b"world".to_vec()), - ], - CompressionLevel::Default, - ) - .await - .expect("zip creation should succeed"); - - let entries = extract_zip_simple(&zip_path, &extract_path) - .await - .expect("zip extraction should succeed"); - assert_eq!(entries.len(), 2); - assert_eq!( - fs::read(extract_path.join("nested/hello.txt")) - .await - .expect("nested zip entry should be extracted"), - b"hello" - ); - assert_eq!( - fs::read(extract_path.join("world.txt")) - .await - .expect("root zip entry should be extracted"), - b"world" - ); - assert!(entries.iter().all(|entry| entry.archive_kind == ArchiveKind::Zip)); - assert!(entries.iter().all(|entry| entry.format == ArchiveFormat::Zip)); - } - - #[tokio::test] - async fn test_create_zip_with_directory_entries_and_extract_directory_scenarios() { - let temp = tempdir().expect("tempdir should be created"); - let explicit_zip_path = temp.path().join("explicit-directories.zip"); - let explicit_extract_path = temp.path().join("explicit-extract"); - let auto_zip_path = temp.path().join("auto-directories.zip"); - let auto_extract_path = temp.path().join("auto-extract"); - - create_zip_with_options( - &explicit_zip_path, - vec![ - ("nested/".to_string(), Vec::new()), - ("nested/deeper/".to_string(), Vec::new()), - ], - ZipWriteOptions { - compression_level: CompressionLevel::Default, - create_directory_entries: false, - }, - ) - .await - .expect("zip creation with explicit directory entries should succeed"); - - let explicit_entries = extract_zip_with_limits(&explicit_zip_path, &explicit_extract_path, ArchiveLimits::default()) - .await - .expect("zip extraction with explicit directory entries should succeed"); - - assert!( - explicit_entries - .iter() - .any(|entry| entry.name.trim_end_matches('/') == "nested" && entry.is_dir) - ); - assert!( - explicit_entries - .iter() - .any(|entry| entry.name.trim_end_matches('/') == "nested/deeper" && entry.is_dir) - ); - assert!( - fs::metadata(explicit_extract_path.join("nested")) - .await - .expect("nested directory should exist") - .is_dir() - ); - assert!( - fs::metadata(explicit_extract_path.join("nested/deeper")) - .await - .expect("nested deeper directory should exist") - .is_dir() - ); - - create_zip_with_options( - &auto_zip_path, - vec![("nested/deeper/file.txt".to_string(), b"hello".to_vec())], - ZipWriteOptions { - compression_level: CompressionLevel::Default, - create_directory_entries: true, - }, - ) - .await - .expect("zip creation with automatic directory entries should succeed"); - - let entries = extract_zip_with_limits(&auto_zip_path, &auto_extract_path, ArchiveLimits::default()) - .await - .expect("zip extraction with automatic directory entries should succeed"); - - assert!( - entries - .iter() - .any(|entry| entry.name.trim_end_matches('/') == "nested" && entry.is_dir) - ); - assert!( - entries - .iter() - .any(|entry| entry.name.trim_end_matches('/') == "nested/deeper" && entry.is_dir) - ); - assert!( - fs::metadata(auto_extract_path.join("nested")) - .await - .expect("nested directory should exist") - .is_dir() - ); - assert!( - fs::metadata(auto_extract_path.join("nested/deeper")) - .await - .expect("nested deeper directory should exist") - .is_dir() - ); - assert_eq!( - fs::read(auto_extract_path.join("nested/deeper/file.txt")) - .await - .expect("nested file should be extracted"), - b"hello" - ); - } - - #[tokio::test] - async fn test_create_zip_with_lots_of_small_files_round_trip() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("many-small-files.zip"); - let extract_path = temp.path().join("extract"); - let files = (0..32) - .map(|index| (format!("batch/file-{index}.txt"), format!("payload-{index}").into_bytes())) - .collect::>(); - - create_zip_with_options( - &zip_path, - files.clone(), - ZipWriteOptions { - compression_level: CompressionLevel::Default, - create_directory_entries: true, - }, - ) - .await - .expect("zip creation for many small files should succeed"); - - let entries = extract_zip_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("zip extraction for many small files should succeed"); - - assert!(entries.len() >= files.len()); - for (path, expected) in files { - assert_eq!( - fs::read(extract_path.join(path)) - .await - .expect("small file should be extracted"), - expected - ); - } - } - - #[tokio::test] - async fn test_extract_zip_to_path_with_limits_returns_summary() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("summary.zip"); - let extract_path = temp.path().join("extract"); - - create_zip_with_options( - &zip_path, - vec![ - ("nested/".to_string(), Vec::new()), - ("nested/hello.txt".to_string(), b"hello".to_vec()), - ("world.txt".to_string(), b"world".to_vec()), - ], - ZipWriteOptions { - compression_level: CompressionLevel::Default, - create_directory_entries: false, - }, - ) - .await - .expect("zip creation for summary should succeed"); - - let summary = extract_zip_to_path_with_limits(&zip_path, &extract_path, ArchiveLimits::default()) - .await - .expect("zip extract summary should succeed"); - - assert_eq!(summary.entry_count, 3); - assert_eq!(summary.directory_count, 1); - assert_eq!(summary.file_count, 2); - assert_eq!(summary.total_unpacked_size, 10); - assert_eq!( - fs::read(extract_path.join("nested/hello.txt")) - .await - .expect("nested file should be extracted"), - b"hello" - ); - assert_eq!( - fs::read(extract_path.join("world.txt")) - .await - .expect("world file should be extracted"), - b"world" - ); - } - - #[tokio::test] - async fn test_zip_helper_exposes_stored_vs_deflated_metadata() { - let temp = tempdir().expect("tempdir should be created"); - let stored_zip_path = temp.path().join("stored.zip"); - let deflated_zip_path = temp.path().join("deflated.zip"); - let stored_extract_path = temp.path().join("stored-extract"); - let deflated_extract_path = temp.path().join("deflated-extract"); - let payload = b"compressible-content-".repeat(64); - - create_zip_with_options( - &stored_zip_path, - vec![("payload.txt".to_string(), payload.clone())], - ZipWriteOptions { - compression_level: CompressionLevel::Fastest, - create_directory_entries: false, - }, - ) - .await - .expect("stored zip creation should succeed"); - - create_zip_with_options( - &deflated_zip_path, - vec![("payload.txt".to_string(), payload.clone())], - ZipWriteOptions { - compression_level: CompressionLevel::Best, - create_directory_entries: false, - }, - ) - .await - .expect("deflated zip creation should succeed"); - - let stored_entries = extract_zip_with_limits(&stored_zip_path, &stored_extract_path, ArchiveLimits::default()) - .await - .expect("stored zip extraction should succeed"); - let deflated_entries = extract_zip_with_limits(&deflated_zip_path, &deflated_extract_path, ArchiveLimits::default()) - .await - .expect("deflated zip extraction should succeed"); - - assert_eq!(stored_entries[0].compression_method, "Stored"); - assert_eq!(deflated_entries[0].compression_method, "Deflated"); - assert_eq!( - fs::read(stored_extract_path.join("payload.txt")) - .await - .expect("stored payload should be extracted"), - payload - ); - assert_eq!( - fs::read(deflated_extract_path.join("payload.txt")) - .await - .expect("deflated payload should be extracted"), - payload - ); - assert!(deflated_entries[0].compressed_size <= stored_entries[0].compressed_size); - } - - #[tokio::test] - async fn test_create_zip_rejects_unsafe_entry_name() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("archive.zip"); - - let err = create_zip_simple( - &zip_path, - vec![("../escape.txt".to_string(), b"escape".to_vec())], - CompressionLevel::Default, - ) - .await - .unwrap_err(); - - assert!(matches!(err, ZipError::UnsafeEntryPath(path) if path == "../escape.txt")); - } - - #[tokio::test] - async fn test_extract_zip_rejects_parent_traversal_entry_and_writes_nothing_outside_target() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("traversal.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("safe.txt", b"safe"), ("../escape.txt", b"escape")]) - .await - .expect("zip fixture should be created"); - - let err = extract_zip_simple(&zip_path, &extract_path) - .await - .expect_err("zip entry with parent traversal should be rejected"); - - assert!(matches!(err, ZipError::UnsafeEntryPath(path) if path == "../escape.txt")); - assert!( - fs::metadata(temp.path().join("escape.txt")).await.is_err(), - "traversal entry must not be written outside the extraction target" - ); - assert!( - fs::metadata(extract_path.join("escape.txt")).await.is_err(), - "traversal entry must not be written inside the extraction target either" - ); - } - - #[tokio::test] - async fn test_extract_zip_rejects_small_entry_with_corrupted_crc() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("corrupt-crc.zip"); - let extract_path = temp.path().join("extract"); - - // A stored entry well under the small-entry fast-path limit so extraction takes - // the in-memory buffer path rather than the streaming `io::copy` path. - let content = b"small-entry-crc-payload"; - assert!((content.len() as u64) <= SMALL_ZIP_EXTRACT_FAST_PATH_LIMIT); - build_zip_file_with_entries(&zip_path, &[("small.txt", content)]) - .await - .expect("zip fixture should be created"); - - // Corrupt the stored entry data so its bytes no longer match the recorded CRC32. - let mut raw = fs::read(&zip_path).await.expect("zip fixture should be readable"); - let offset = raw - .windows(content.len()) - .position(|window| window == content) - .expect("stored entry data should be present in the zip"); - raw[offset] ^= 0xFF; - fs::write(&zip_path, &raw).await.expect("corrupted zip should be writable"); - - let err = extract_zip_simple(&zip_path, &extract_path) - .await - .expect_err("small entry with a corrupted CRC should be rejected"); - - assert!( - matches!(err, ZipError::Io(ref e) if e.kind() == std::io::ErrorKind::InvalidData), - "expected an InvalidData checksum error, got {err:?}" - ); - } - - #[tokio::test] - async fn test_extract_zip_allows_entry_count_exactly_at_limit() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("at-limit.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("one.txt", b"1"), ("two.txt", b"2")]) - .await - .expect("zip fixture should be created"); - - let entries = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_entries: 2, - ..ArchiveLimits::default() - }, - ) - .await - .expect("entry count exactly at the limit should extract successfully"); - - assert_eq!(entries.len(), 2); - } - - #[tokio::test] - async fn test_extract_zip_allows_total_unpacked_size_exactly_at_limit() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("total-at-limit.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("one.txt", b"12345"), ("two.txt", b"67890")]) - .await - .expect("zip fixture should be created"); - - let entries = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_total_unpacked_size: 10, - ..ArchiveLimits::default() - }, - ) - .await - .expect("total unpacked size exactly at the limit should extract successfully"); - - assert_eq!(entries.len(), 2); - assert_eq!( - fs::read(extract_path.join("two.txt")) - .await - .expect("second entry should be extracted"), - b"67890" - ); - } - - #[tokio::test] - async fn test_extract_zip_rejects_too_many_entries() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("too-many.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("one.txt", b"1"), ("two.txt", b"2")]) - .await - .expect("zip fixture should be created"); - - let err = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_entries: 1, - ..ArchiveLimits::default() - }, - ) - .await - .expect_err("zip entry count limit should fail"); - - assert!(matches!(err, ZipError::EntryCountLimitExceeded { count: 2, limit: 1 })); - } - - #[tokio::test] - async fn test_extract_zip_rejects_oversized_entry() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("oversized.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("big.txt", b"hello world")]) - .await - .expect("zip fixture should be created"); - - let err = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_entry_size: 4, - ..ArchiveLimits::default() - }, - ) - .await - .expect_err("zip entry size limit should fail"); - - assert!(matches!( - err, - ZipError::EntrySizeLimitExceeded { - path, - size: 11, - limit: 4, - } if path == "big.txt" - )); - } - - #[tokio::test] - async fn test_extract_zip_rejects_total_unpacked_size_limit() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("too-large-total.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("one.txt", b"12345"), ("two.txt", b"67890")]) - .await - .expect("zip fixture should be created"); - - let err = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_total_unpacked_size: 9, - ..ArchiveLimits::default() - }, - ) - .await - .expect_err("zip total size limit should fail"); - - assert!(matches!(err, ZipError::TotalUnpackedSizeLimitExceeded { size: 10, limit: 9 })); - } - - #[tokio::test] - async fn test_extract_zip_rejects_entry_path_length_limit() { - let temp = tempdir().expect("tempdir should be created"); - let zip_path = temp.path().join("long-path.zip"); - let extract_path = temp.path().join("extract"); - - build_zip_file_with_entries(&zip_path, &[("nested/hello.txt", b"hello")]) - .await - .expect("zip fixture should be created"); - - let err = extract_zip_with_limits( - &zip_path, - &extract_path, - ArchiveLimits { - max_path_length: 5, - ..ArchiveLimits::default() - }, - ) - .await - .expect_err("zip path length limit should fail"); - - assert!(matches!( - err, - ZipError::EntryPathTooLong { path, limit: 5, .. } if path == "nested/hello.txt" - )); - } - - #[tokio::test] - async fn test_decompress_file_round_trip() { - let temp = tempdir().expect("tempdir should be created"); - let input_path = temp.path().join("payload.txt.gz"); - let output_path = temp.path().join("payload.txt"); - let compressed = Compressor::new(CompressionFormat::Gzip) - .compress(b"payload") - .await - .expect("gzip compress should succeed"); - fs::write(&input_path, compressed) - .await - .expect("compressed input file should be written"); - - Decompressor::auto_detect(&input_path) - .decompress_file(&input_path, &output_path) - .await - .expect("gzip file decompress should succeed"); - - assert_eq!( - fs::read(&output_path) - .await - .expect("decompressed output file should be readable"), - b"payload" - ); - } } From 9852e53b4ccaa9052ccfbfd29e1bef45ab009a82 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 18:28:13 +0800 Subject: [PATCH 15/24] test(lifecycle,scanner): drop 47 no-op #[serial] markers (backlog#1846 T1) (#6213) Second batch of the #[serial] sweep started in #6209. nextest is the repository's authoritative runner and executes every test in its own process, so serial_test's in-process mutex cannot serialize tests against each other -- docs/testing/README.md documents this, and the mechanism that actually serializes across the process boundary is a .config/nextest.toml [test-groups] entry with max-threads = 1. Unlike the first batch (e2e, process-isolated by construction), these are in-crate unit tests that could genuinely share process state under the `cargo test` fallback runner, where #[serial] IS still effective. Every marker was therefore reviewed individually and removed only where the test provably touches neither the process environment nor a process-global. Removed (47, pure deletions, no test bodies touched): crates/lifecycle/src/core.rs 35 crates/scanner/src/scanner.rs 12 The lifecycle removals are all validate_* / filter_rules_* / has_active_rules_* / noncurrent_versions_expiration_limit_* tests that build a local BucketLifecycleConfiguration and call a &self method walking only that value. The scanner removals are pure duration arithmetic (randomized_cycle_delay_for, initial_scanner_delay_for with an explicit Some(secs), the bitrot-disabled early return of scanner_clean_idle_max_interval) and background_heal_info_for_scan_complete / _for_scan_result field comparisons over locally built values. Retained deliberately -- see the PR body for the full list and reasons: crates/ecstore/src/bucket/lifecycle/bucket_lifecycle_ops.rs 101 (all) crates/lifecycle/src/core.rs 44 crates/scanner/src/scanner.rs 53 bucket_lifecycle_ops.rs keeps every marker: its test module caches a process-wide `static STALE_MULTIPART_TEST_ENV: OnceLock<(Vec, Arc)>`, and its own reregister_env_local_disks helper documents in-tree that sibling #[serial] tests reset and reshape the shared local-disk registry for each other. That sharing is real, so the markers stay. No test was renamed, added, or deleted; no reserved migration-gate name substring is affected; no .config/nextest.toml entry references any of the 47 removed tests. --- crates/lifecycle/src/core.rs | 35 ----------------------------------- crates/scanner/src/scanner.rs | 12 ------------ 2 files changed, 47 deletions(-) diff --git a/crates/lifecycle/src/core.rs b/crates/lifecycle/src/core.rs index ff7d15505..0fe52b9df 100644 --- a/crates/lifecycle/src/core.rs +++ b/crates/lifecycle/src/core.rs @@ -1199,7 +1199,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_zero_expiration_days() { // S3 compatibility: Expiration.Days must be a positive integer (>= 1). AWS and // the ceph s3-tests `test_lifecycle_expiration_days0` case reject Days == 0 with @@ -1233,7 +1232,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_negative_expiration_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1263,7 +1261,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_accepts_positive_expiration_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1290,7 +1287,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_accepts_one_day_boundary_values() { // Pin the exact >= 1 boundary: a value of 1 is the smallest legal positive // integer and must be accepted for every day-count field tightened for S3 @@ -1325,7 +1321,6 @@ mod tests { } #[tokio::test] - #[serial] async fn has_active_rules_accepts_zero_day_expiration() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1350,7 +1345,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_zero_noncurrent_expiration_days() { // S3 compatibility: NoncurrentVersionExpiration.NoncurrentDays must be a positive // integer (>= 1); AWS rejects 0 with InvalidArgument. @@ -1382,7 +1376,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_negative_noncurrent_expiration_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1412,7 +1405,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_accepts_abort_incomplete_multipart_upload_only_rule() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1438,7 +1430,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_zero_abort_incomplete_multipart_upload_days() { // S3 compatibility: AbortIncompleteMultipartUpload.DaysAfterInitiation must be a // positive integer (>= 1); AWS rejects 0 with InvalidArgument. @@ -1469,7 +1460,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_missing_abort_incomplete_multipart_upload_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1495,7 +1485,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_negative_abort_incomplete_multipart_upload_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1556,7 +1545,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_non_midnight_expiration_date() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1638,7 +1626,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_accepts_multiple_rules_without_ids() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1682,7 +1669,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_rule_id_too_long() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1709,7 +1695,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_duplicate_rule_ids() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1752,7 +1737,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_transition_without_storage_class() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1780,7 +1764,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_transition_without_date_or_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -1808,7 +1791,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_noncurrent_transition_without_days() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -2365,7 +2347,6 @@ mod tests { } #[tokio::test] - #[serial] async fn noncurrent_versions_expiration_limit_returns_configured_limits() { let lc = Arc::new(BucketLifecycleConfiguration { expiry_updated_at: None, @@ -2456,7 +2437,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_invalid_status_case_sensitive() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -2483,7 +2463,6 @@ mod tests { } #[tokio::test] - #[serial] async fn filter_rules_respects_filter_prefix() { let filter = LifecycleRuleFilter { prefix: Some("prefix".to_string()), @@ -2528,7 +2507,6 @@ mod tests { } #[tokio::test] - #[serial] async fn filter_rules_respects_filter_and_prefix() { let and = s3s::dto::LifecycleRuleAndOperator { prefix: Some("prefix".to_string()), @@ -2578,7 +2556,6 @@ mod tests { } #[tokio::test] - #[serial] async fn filter_rules_respects_filter_tag() { let filter = LifecycleRuleFilter { tag: Some(s3s::dto::Tag { @@ -2632,7 +2609,6 @@ mod tests { } #[tokio::test] - #[serial] async fn filter_rules_respects_filter_and_tags() { let filter = LifecycleRuleFilter { and: Some(s3s::dto::LifecycleRuleAndOperator { @@ -3086,7 +3062,6 @@ mod tests { // --- TASK-002 tests: Object Lock + ExpiredObjectDeleteMarker compatibility --- #[tokio::test] - #[serial] async fn validate_allows_expired_object_delete_marker_on_locked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3118,7 +3093,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_expired_object_delete_marker_on_unlocked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3146,7 +3120,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_non_delete_marker_expiration_on_locked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3179,7 +3152,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_del_marker_expiration_on_locked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3210,7 +3182,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_rejects_zero_day_del_marker_expiration_on_locked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3594,7 +3565,6 @@ mod tests { // --- TASK-007 tests: Legacy Prefix/Filter conflict --- #[tokio::test] - #[serial] async fn validate_rejects_prefix_and_filter_both_present() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3623,7 +3593,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_prefix_without_filter() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3650,7 +3619,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_filter_without_prefix() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3680,7 +3648,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_empty_prefix_with_filter() { // Empty prefix should be treated as "not set" let lc = BucketLifecycleConfiguration { @@ -3713,7 +3680,6 @@ mod tests { // --- TASK-004 tests: ExpiredObjectAllVersions --- #[tokio::test] - #[serial] async fn validate_rejects_expired_object_all_versions_on_locked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, @@ -3745,7 +3711,6 @@ mod tests { } #[tokio::test] - #[serial] async fn validate_allows_expired_object_all_versions_on_unlocked_bucket() { let lc = BucketLifecycleConfiguration { expiry_updated_at: None, diff --git a/crates/scanner/src/scanner.rs b/crates/scanner/src/scanner.rs index 11023eaac..fe743a1c6 100644 --- a/crates/scanner/src/scanner.rs +++ b/crates/scanner/src/scanner.rs @@ -4574,7 +4574,6 @@ mod tests { } #[test] - #[serial] fn test_randomized_cycle_delay_keeps_configured_start_delay() { // 120s with ±10% jitter should stay clearly above the historic 30s cap. let delay = randomized_cycle_delay_for(Duration::from_secs(120)); @@ -4593,7 +4592,6 @@ mod tests { } #[test] - #[serial] fn test_initial_scanner_delay_uses_configured_start_delay() { let delay = initial_scanner_delay_for(Some(120)); assert!(delay >= Duration::from_secs(108)); @@ -4613,14 +4611,12 @@ mod tests { } #[test] - #[serial] fn test_initial_scanner_delay_skips_for_cold_usage_cache_with_buckets() { let delay = initial_scanner_delay_for_startup(Some(120), true, true, false); assert_eq!(delay, Duration::ZERO); } #[test] - #[serial] fn test_initial_scanner_delay_keeps_configured_delay_for_warm_usage_cache_no_replication() { let delay = initial_scanner_delay_for_startup(Some(120), false, true, false); assert!(delay >= Duration::from_secs(108)); @@ -4628,14 +4624,12 @@ mod tests { } #[test] - #[serial] fn test_initial_scanner_delay_skips_for_cold_usage_cache_without_buckets() { let delay = initial_scanner_delay_for_startup(Some(120), true, false, false); assert_eq!(delay, Duration::ZERO); } #[test] - #[serial] fn test_initial_scanner_delay_skips_for_active_replication_warm_cache() { // Warm cache + active replication rules → skip startup delay so that FAILED-status objects // from a crash are healed on the first cycle, not after a 27-33 min sleep. @@ -4644,7 +4638,6 @@ mod tests { } #[test] - #[serial] fn test_initial_scanner_delay_keeps_delay_for_replication_without_buckets() { // Active replication but no buckets → no objects to scan, keep normal delay. let delay = initial_scanner_delay_for_startup(Some(120), false, false, true); @@ -7399,7 +7392,6 @@ mod tests { } #[test] - #[serial] fn clean_idle_cap_allows_policy_max_when_bitrot_is_disabled() { let config = ScannerRuntimeConfig { bitrot_cycle: None, @@ -7515,7 +7507,6 @@ mod tests { } #[test] - #[serial] fn test_randomized_cycle_delay_handles_small_start_delay() { // 0 is treated as minimum 1 second before jitter, with lower bound preserved. let delay = randomized_cycle_delay_for(Duration::from_secs(0)); @@ -8174,7 +8165,6 @@ mod tests { } #[test] - #[serial] fn test_background_heal_info_for_scan_complete_marks_deep_idle() { let started_at = Utc::now(); let info = BackgroundHealInfo { @@ -8192,7 +8182,6 @@ mod tests { } #[test] - #[serial] fn test_background_heal_info_for_scan_complete_leaves_normal_scan_unchanged() { let info = BackgroundHealInfo { bitrot_start_time: Some(Utc::now()), @@ -8204,7 +8193,6 @@ mod tests { } #[test] - #[serial] fn test_background_heal_info_for_failed_scan_preserves_deep_mode() { let info = BackgroundHealInfo { bitrot_start_time: Some(Utc::now()), From 8a3c66e655c2baabcd1a27171947761093f4a406 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 18:28:46 +0800 Subject: [PATCH 16/24] test(admin): replace a source-text guard with a behavior test (#6212) --- crates/test-utils/src/lib.rs | 10 ++- rustfs/src/admin/handlers/system.rs | 78 +++++++++++++++++++++ rustfs/src/admin/route_registration_test.rs | 17 ----- rustfs/src/admin/storage_api.rs | 4 ++ 4 files changed, 91 insertions(+), 18 deletions(-) diff --git a/crates/test-utils/src/lib.rs b/crates/test-utils/src/lib.rs index 46f215a9d..59b4f3e0c 100644 --- a/crates/test-utils/src/lib.rs +++ b/crates/test-utils/src/lib.rs @@ -92,6 +92,13 @@ pub struct TestECStoreEnv { /// `init_local_disks` + `ECStore::new` on `127.0.0.1:0` (random port keeps /// nextest's process-per-test parallelism safe). pub ecstore: Arc, + /// The single-pool, single-set topology the store was built from. + /// + /// The bootstrap does **not** publish it on the instance context (server + /// startup is what calls `set_endpoints`, and that write is once-only), so + /// a test that needs `get_global_endpoints` to resolve — admin server-info + /// and other topology readers — publishes this value itself. + pub endpoint_pools: EndpointServerPools, } impl TestECStoreEnv { @@ -234,7 +241,7 @@ impl TestECStoreEnvBuilder { // Port 0 keeps ECStore-backed integration binaries parallel-safe under // nextest: no fixed peer port is ever shared between test processes. let server_addr: std::net::SocketAddr = "127.0.0.1:0".parse().expect("parse test addr"); - let ecstore = ECStore::new(server_addr, endpoint_pools, CancellationToken::new()) + let ecstore = ECStore::new(server_addr, endpoint_pools.clone(), CancellationToken::new()) .await .expect("build test ECStore"); @@ -254,6 +261,7 @@ impl TestECStoreEnvBuilder { temp_root, disk_paths, ecstore, + endpoint_pools, } } } diff --git a/rustfs/src/admin/handlers/system.rs b/rustfs/src/admin/handlers/system.rs index 0909568e5..9df481c11 100644 --- a/rustfs/src/admin/handlers/system.rs +++ b/rustfs/src/admin/handlers/system.rs @@ -1541,6 +1541,84 @@ mod tests { assert_eq!(error.code(), &S3ErrorCode::InvalidRequest); } + /// `ServerInfoHandler` must answer an authorized admin request with the + /// per-pool erasure-set topology (rustfs/backlog#1839). That map is only + /// filled when the server-info query is issued with pools included, so a + /// handler that stopped asking for them would still return 200 with an + /// empty `pools` object instead of failing. + #[tokio::test] + #[serial_test::serial] + async fn server_info_response_carries_pool_topology() { + use crate::admin::runtime_sources::{AppContext, publish_test_app_context}; + use crate::admin::storage_api::runtime::bootstrap_ctx; + use http_body_util::BodyExt as _; + use rustfs_iam::store::{Store as _, object::IAM_CONFIG_PREFIX}; + use std::sync::Arc; + + const ROOT_ACCESS_KEY: &str = "SERVERINFOROOTACCESSKEY"; + const ROOT_SECRET_KEY: &str = "serverInfoRootSecret123"; + + let _ = rustfs_credentials::init_global_action_credentials( + Some(ROOT_ACCESS_KEY.to_string()), + Some(ROOT_SECRET_KEY.to_string()), + ); + + let env = rustfs_test_utils::TestECStoreEnv::builder() + .prefix("admin_server_info_pools") + .disk_count(1) + .init_bucket_metadata(false) + .build() + .await; + // Server startup owns this write in production; the test bootstrap + // stops short of it, and without a topology the server-info query + // returns before it ever looks at drives. + bootstrap_ctx().set_endpoints(env.endpoint_pools.clone()); + rustfs_iam::store::object::ObjectStore::new(Arc::clone(&env.ecstore)) + .save_iam_config(serde_json::json!({"version": 1}), format!("{}/format.json", *IAM_CONFIG_PREFIX)) + .await + .expect("seed IAM format"); + let iam = rustfs_iam::build_iam_sys(Arc::clone(&env.ecstore)) + .await + .expect("build test IAM"); + publish_test_app_context(Arc::new(AppContext::with_default_interfaces( + Arc::clone(&env.ecstore), + iam, + Arc::new(rustfs_kms::KmsServiceManager::new()), + ))); + + let request = S3Request { + input: Body::empty(), + method: Method::GET, + uri: Uri::from_static("/rustfs/admin/v3/info"), + headers: HeaderMap::new(), + extensions: Extensions::new(), + credentials: Some(s3s::auth::Credentials { + access_key: ROOT_ACCESS_KEY.to_string(), + secret_key: s3s::auth::SecretKey::from(ROOT_SECRET_KEY.to_string()), + }), + region: None, + service: None, + trailing_headers: None, + }; + + let (status, body) = super::ServerInfoHandler {} + .call(request, Params::new()) + .await + .expect("root admin credentials must be served server info") + .output; + assert_eq!(status, hyper::StatusCode::OK); + + let bytes = body.collect().await.expect("server info body should read").to_bytes(); + let payload: serde_json::Value = serde_json::from_slice(&bytes).expect("server info must be json"); + let pools = payload["info"]["pools"] + .as_object() + .expect("server info must carry a pools object"); + assert!( + pools.contains_key("0"), + "server info must report the erasure-set topology of pool 0, got {pools:?}" + ); + } + /// Authorization denial for this exact action is pinned to AccessDenied by /// `crate::admin::auth::tests::non_admin_credential_is_denied`. #[test] diff --git a/rustfs/src/admin/route_registration_test.rs b/rustfs/src/admin/route_registration_test.rs index e81f13d17..e48e09c94 100644 --- a/rustfs/src/admin/route_registration_test.rs +++ b/rustfs/src/admin/route_registration_test.rs @@ -1412,23 +1412,6 @@ fn test_health_routes_not_registered_when_disabled_by_env() { }); } -#[test] -fn test_phase5_admin_info_contract() { - let system_src = include_str!("handlers/system.rs"); - - let server_info_impl_marker = "impl Operation for ServerInfoHandler"; - let server_info_impl_start = system_src - .find(server_info_impl_marker) - .expect("Expected impl Operation for ServerInfoHandler in handlers/system.rs"); - let server_info_impl_block = &system_src[server_info_impl_start..]; - - assert!( - server_info_impl_block.contains("default_admin_usecase()") - && server_info_impl_block.contains("execute_query_server_info(QueryServerInfoRequest { include_pools: true })"), - "admin server info path must be served through admin runtime-source DefaultAdminUsecase::execute_query_server_info" - ); -} - fn extract_block_between_markers<'a>(src: &'a str, start_marker: &str, end_marker: &str) -> &'a str { let start = src .find(start_marker) diff --git a/rustfs/src/admin/storage_api.rs b/rustfs/src/admin/storage_api.rs index ec89f5e4e..dbf3a81df 100644 --- a/rustfs/src/admin/storage_api.rs +++ b/rustfs/src/admin/storage_api.rs @@ -948,6 +948,10 @@ pub(crate) mod runtime { #[cfg(test)] pub(crate) use super::{Endpoint, Endpoints, PoolEndpoints}; + /// Test-only: the process instance context, so a handler test can publish + /// the endpoint topology that server startup normally installs. + #[cfg(test)] + pub(crate) use crate::storage::storage_api::ecstore_runtime::bootstrap_ctx; } pub(crate) mod s3 { From 127b662f3f00f6c985a3e1203faecf90fc7fcf1a Mon Sep 17 00:00:00 2001 From: houseme Date: Tue, 18 Aug 2026 19:53:06 +0800 Subject: [PATCH 17/24] feat(app): add opt-in small GET body once path (#6216) Use the merged s3s single-chunk StreamingBlob support for exact-length materialized GET bodies when RUSTFS_GET_SMALL_BODY_ONCE_ENABLE is enabled. Keep the default path unchanged and fall back to the guarded MemoryTrackedBytesStream on length mismatch. Co-authored-by: heihutu --- Cargo.lock | 5 +- Cargo.toml | 2 +- rustfs/src/app/object_usecase.rs | 84 +++++++++++++++++++++++++++++++- 3 files changed, 87 insertions(+), 4 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index 82f66641e..db75a7992 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -10678,7 +10678,7 @@ checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f" [[package]] name = "s3s" version = "0.14.1" -source = "git+https://github.com/rustfs/s3s.git?rev=d7028511a53f69d41ed3c69f36899f9b1aede647#d7028511a53f69d41ed3c69f36899f9b1aede647" +source = "git+https://github.com/rustfs/s3s.git?rev=ff8106f46098b60fce8ee88518f34974fbfbec4b#ff8106f46098b60fce8ee88518f34974fbfbec4b" dependencies = [ "arc-swap", "arrayvec", @@ -10706,6 +10706,7 @@ dependencies = [ "numeric_cast", "pin-project-lite", "quick-xml", + "regex", "serde", "serde_json", "serde_urlencoded", @@ -11809,7 +11810,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd" dependencies = [ "fastrand", - "getrandom 0.4.3", + "getrandom 0.3.4", "once_cell", "rustix", "windows-sys 0.61.2", diff --git a/Cargo.toml b/Cargo.toml index ae9b497e0..f2080a15f 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -290,7 +290,7 @@ rustify = { version = "0.7", default-features = false } rustix = { version = "1.1.4" } rust-embed = { version = "8.12.0" } rustc-hash = { version = "2.1.3" } -s3s = { git = "https://github.com/rustfs/s3s.git", rev = "d7028511a53f69d41ed3c69f36899f9b1aede647" } +s3s = { git = "https://github.com/rustfs/s3s.git", rev = "ff8106f46098b60fce8ee88518f34974fbfbec4b" } serial_test = "4.0.1" shadow-rs = { default-features = false, version = "2.0.0" } siphasher = "1.0.3" diff --git a/rustfs/src/app/object_usecase.rs b/rustfs/src/app/object_usecase.rs index 1bb337c69..bcce2b705 100644 --- a/rustfs/src/app/object_usecase.rs +++ b/rustfs/src/app/object_usecase.rs @@ -782,6 +782,7 @@ const MID_BODY_READER_STREAM_BUFFER_THRESHOLD_BYTES: i64 = MI_B as i64; const ENV_RUSTFS_GET_SEEK_BUFFER_ENABLE: &str = "RUSTFS_GET_SEEK_BUFFER_ENABLE"; const ENV_RUSTFS_GET_READER_STREAM_BUFFER_SIZE: &str = "RUSTFS_GET_READER_STREAM_BUFFER_SIZE"; const ENV_RUSTFS_GET_OUTPUT_HANDOFF_ATTRIBUTION_ENABLE: &str = "RUSTFS_GET_OUTPUT_HANDOFF_ATTRIBUTION_ENABLE"; +const ENV_RUSTFS_GET_SMALL_BODY_ONCE_ENABLE: &str = "RUSTFS_GET_SMALL_BODY_ONCE_ENABLE"; const GET_READER_STREAM_BUFFER_SOURCE_SELECTED: &str = "selected"; const GET_READER_STREAM_BUFFER_SOURCE_ENV_OVERRIDE: &str = "env_override"; const GET_READER_STREAM_POLL_PENDING: &str = "pending"; @@ -813,6 +814,18 @@ fn is_get_output_handoff_attribution_enabled() -> bool { *ENABLED.get_or_init(|| rustfs_utils::get_env_bool(ENV_RUSTFS_GET_OUTPUT_HANDOFF_ATTRIBUTION_ENABLE, false)) } +fn is_get_small_body_once_enabled() -> bool { + #[cfg(test)] + { + rustfs_utils::get_env_bool(ENV_RUSTFS_GET_SMALL_BODY_ONCE_ENABLE, false) + } + #[cfg(not(test))] + { + static ENABLED: OnceLock = OnceLock::new(); + *ENABLED.get_or_init(|| rustfs_utils::get_env_bool(ENV_RUSTFS_GET_SMALL_BODY_ONCE_ENABLE, false)) + } +} + fn is_get_seek_buffer_enabled() -> bool { static ENABLED: OnceLock = OnceLock::new(); *ENABLED.get_or_init(|| rustfs_utils::get_env_bool(ENV_RUSTFS_GET_SEEK_BUFFER_ENABLE, false)) @@ -932,6 +945,30 @@ struct MemoryTrackedBytesStream { lifecycle: GetObjectBodyLifecycle, } +struct MemoryOnceBodyOwner { + bytes: Bytes, + _guard: Option, + // Body::Once has no poll hook, so this opt-in path only holds the request + // guard until the bytes are dropped; the result status remains unknown. + _lifecycle: GetObjectBodyLifecycle, +} + +impl MemoryOnceBodyOwner { + fn new(bytes: Bytes, guard: Option, lifecycle: GetObjectBodyLifecycle) -> Self { + Self { + bytes, + _guard: guard, + _lifecycle: lifecycle, + } + } +} + +impl AsRef<[u8]> for MemoryOnceBodyOwner { + fn as_ref(&self) -> &[u8] { + self.bytes.as_ref() + } +} + #[derive(Default)] struct GetObjectBodyLifecycle { request_guard: Option, @@ -4160,7 +4197,12 @@ impl DefaultObjectUsecase { let bytes_len = bytes.len(); let guard = rustfs_io_metrics::track_get_object_buffered_bytes(bytes_len); let remaining = usize::try_from(response_content_length.max(0)).unwrap_or(usize::MAX); - let blob = StreamingBlob::new(MemoryTrackedBytesStream::new(bytes, remaining, source, guard, lifecycle)); + let blob = if is_get_small_body_once_enabled() && bytes_len == remaining { + let owner = MemoryOnceBodyOwner::new(bytes, guard, lifecycle); + StreamingBlob::from_bytes(Bytes::from_owner(owner)) + } else { + StreamingBlob::new(MemoryTrackedBytesStream::new(bytes, remaining, source, guard, lifecycle)) + }; if let Some(handoff_start) = handoff_start { rustfs_io_metrics::record_get_object_response_handoff( "single_chunk", @@ -12955,6 +12997,46 @@ mod tests { assert_eq!(blob.remaining_length().exact(), Some(5)); } + #[test] + #[serial_test::serial] + fn memory_blob_once_fast_path_holds_guard_until_bytes_drop() { + temp_env::with_var(ENV_RUSTFS_GET_SMALL_BODY_ONCE_ENABLE, Some("true"), || { + let initial = GetObjectGuard::concurrent_count(); + let guard = GetObjectGuard::new(); + assert_eq!(GetObjectGuard::concurrent_count(), initial + 1); + + let blob = DefaultObjectUsecase::build_memory_bytes_blob( + Bytes::from_static(b"hello"), + 5, + GET_MEMORY_BODY_SOURCE_BUFFERED_BODY, + GetObjectBodyLifecycle::tracked(guard), + ); + let mut body = s3s::Body::from(blob); + let bytes = body.take_bytes().expect("opt-in exact memory body should stay on Body::Once"); + + assert_eq!(bytes, Bytes::from_static(b"hello")); + assert_eq!(GetObjectGuard::concurrent_count(), initial + 1); + drop(bytes); + assert_eq!(GetObjectGuard::concurrent_count(), initial); + }); + } + + #[test] + #[serial_test::serial] + fn memory_blob_once_fast_path_rejects_length_mismatch() { + temp_env::with_var(ENV_RUSTFS_GET_SMALL_BODY_ONCE_ENABLE, Some("true"), || { + let blob = DefaultObjectUsecase::build_memory_bytes_blob( + Bytes::from_static(b"test"), + 5, + GET_MEMORY_BODY_SOURCE_BUFFERED_BODY, + GetObjectBodyLifecycle::disabled(), + ); + let mut body = s3s::Body::from(blob); + + assert!(body.take_bytes().is_none(), "mismatched memory body must keep the guarded stream path"); + }); + } + #[tokio::test] async fn get_object_streaming_reader_times_out_when_body_stalls() { let reader = GetObjectStreamingReader::new( From 27d23b61351a81b1aec1125d7afb49f1fc1f1ee1 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 21:21:34 +0800 Subject: [PATCH 18/24] test(io-metrics): assert record helper emissions; fix census heuristic (#6217) * test(io-metrics): assert lib.rs record helper emissions The 29 assertion-less record_* smoke tests in io-metrics/src/lib.rs called their helpers and checked nothing; because METRICS_ENABLED defaults to false they did not even reach the emission bodies. Replace them with six DebuggingRecorder tests that enable the gate, pin every metric name the helpers own, and pin the derived values, branch selection and label mapping (rustfs/backlog#1836). * test(tooling): anchor assertless-census delegation tokens to name segments --- crates/io-metrics/src/lib.rs | 724 ++++++++++++++++++++++++------- scripts/find_assertless_tests.py | 19 +- 2 files changed, 578 insertions(+), 165 deletions(-) diff --git a/crates/io-metrics/src/lib.rs b/crates/io-metrics/src/lib.rs index 86b38ed5a..785cea246 100644 --- a/crates/io-metrics/src/lib.rs +++ b/crates/io-metrics/src/lib.rs @@ -2633,36 +2633,164 @@ mod tests { use std::sync::{Arc, Barrier, Mutex}; // Serialize tests that mutate the process-global PUT_STAGE_METRICS_ENABLED flag. - static METRICS_FLAG_LOCK: Mutex<()> = Mutex::new(()); + pub(crate) static METRICS_FLAG_LOCK: Mutex<()> = Mutex::new(()); - #[test] - fn test_record_zero_copy_read() { - record_zero_copy_read(1024, 10.5); - record_memory_copy_saved(1024); - record_zero_copy_fallback("test"); + /// One row of a `DebuggingRecorder` snapshot. + pub(crate) type MetricRow = ( + metrics_util::CompositeKey, + Option, + Option, + DebugValue, + ); + + /// Every metric name present in a snapshot. + pub(crate) fn emitted_names(rows: &[MetricRow]) -> std::collections::HashSet<&str> { + rows.iter().map(|(composite, _, _, _)| composite.key().name()).collect() } - #[test] - fn test_record_bytes_pool_metrics() { - record_bytes_pool_acquire("small", 4096, true); - record_bytes_pool_return("small"); - record_bytes_pool_allocated("small", 4096); - record_bytes_pool_hit_rate("small", 0.85); + /// Counter value for `name`, summed over every label set it was emitted with. + /// `None` means the counter never reached the recorder. + pub(crate) fn counter_total(rows: &[MetricRow], name: &str) -> Option { + let mut total = None; + for (composite, _, _, value) in rows { + if composite.kind() == MetricKind::Counter && composite.key().name() == name { + match value { + DebugValue::Counter(count) => *total.get_or_insert(0) += count, + other => panic!("{name} is registered as a counter but holds {other:?}"), + } + } + } + total } - #[test] - fn test_record_bytespool_acquisition_and_return() { - // Acquisition outcomes - record_bytespool_acquisition("small", "hit"); - record_bytespool_acquisition("medium", "miss"); - record_bytespool_acquisition("large", "hit"); - record_bytespool_acquisition("xlarge", "miss"); + /// Gauge value for `name`. Panics when several label sets carry it, so a caller + /// cannot silently assert on an arbitrary one. + pub(crate) fn gauge_value(rows: &[MetricRow], name: &str) -> Option { + let mut matching = rows + .iter() + .filter(|(composite, _, _, _)| composite.kind() == MetricKind::Gauge && composite.key().name() == name); + let value = matching.next().map(|(_, _, _, value)| match value { + DebugValue::Gauge(value) => value.0, + other => panic!("{name} is registered as a gauge but holds {other:?}"), + }); + assert!( + matching.next().is_none(), + "{name} carries several label sets; assert on the labelled rows instead" + ); + value + } - // Return outcomes - record_bytespool_return("small", "recycled"); - record_bytespool_return("medium", "dropped"); - record_bytespool_return("large", "recycled"); - record_bytespool_return("xlarge", "dropped"); + /// Histogram samples for `name` across every label set, sorted so the assertion + /// does not depend on registry iteration order. + pub(crate) fn histogram_samples(rows: &[MetricRow], name: &str) -> Vec { + let mut samples: Vec = rows + .iter() + .filter(|(composite, _, _, _)| composite.kind() == MetricKind::Histogram && composite.key().name() == name) + .flat_map(|(_, _, _, value)| match value { + DebugValue::Histogram(samples) => samples.iter().map(|sample| sample.0), + other => panic!("{name} is registered as a histogram but holds {other:?}"), + }) + .collect(); + samples.sort_by(f64::total_cmp); + samples + } + + /// Replaces four smoke tests that called the zero-copy and bytes-pool recorders + /// and asserted nothing (rustfs/backlog#1836). The same calls now run against a + /// local recorder: every metric name these helpers own must be emitted, the + /// `from_pool` branch must pick the hit/miss counter, and the derived values + /// (byte totals, the hit rate's percent conversion) must match the inputs. + #[test] + fn zero_copy_and_bytes_pool_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); + + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_zero_copy_read(1024, 10.5); + record_memory_copy_saved(1024); + record_zero_copy_fallback("test"); + record_zero_copy_write(2048, 20.5); + record_zero_copy_write_fallback("test"); + record_bytes_saved(4096); + record_bytes_pool_acquire("small", 4096, true); + record_bytes_pool_acquire("small", 4096, false); + record_bytes_pool_return("small"); + record_bytes_pool_allocated("small", 4096); + record_bytes_pool_hit_rate("small", 0.85); + record_bytespool_acquisition("small", "hit"); + record_bytespool_acquisition("medium", "miss"); + record_bytespool_return("small", "recycled"); + record_bytespool_return("medium", "dropped"); + set_metrics_enabled(false); + }); + + let rows = snapshotter.snapshot().into_vec(); + let names = emitted_names(&rows); + for expected in [ + "rustfs_zero_copy_reads_total", + "rustfs_zero_copy_read_size_bytes", + "rustfs_zero_copy_read_duration_ms", + mmap_copy::READS_TOTAL, + mmap_copy::READ_SIZE_BYTES, + mmap_copy::READ_DURATION_MS, + mmap_copy::BYTES_COPIED_TOTAL, + mmap_copy::FALLBACK_TOTAL, + "rustfs_zero_copy_memory_saved_bytes_total", + "rustfs_zero_copy_fallback_total", + "rustfs_zero_copy_write_total", + "rustfs_zero_copy_write_size_bytes", + "rustfs_zero_copy_write_duration_ms", + buffered_write::WRITES_TOTAL, + buffered_write::WRITE_SIZE_BYTES, + buffered_write::WRITE_DURATION_MS, + buffered_write::BYTES_COPIED_TOTAL, + buffered_write::FALLBACK_TOTAL, + "rustfs_zero_copy_write_fallback_total", + "rustfs_zero_copy_bytes_saved_total", + "rustfs_bytes_pool_acquisitions_total", + "rustfs_bytes_pool_size_bytes", + "rustfs_bytes_pool_hits_total", + "rustfs_bytes_pool_misses_total", + "rustfs_bytes_pool_returns_total", + "rustfs_bytes_pool_allocated_bytes", + "rustfs_bytes_pool_hit_rate", + "rustfs_io_bytespool_acquisition_total", + "rustfs_io_bytespool_return_total", + ] { + assert!(names.contains(expected), "{expected} must be emitted by its record helper"); + } + + assert_eq!( + counter_total(&rows, mmap_copy::BYTES_COPIED_TOTAL), + Some(1024), + "the read helper must count the read size, not the call" + ); + assert_eq!( + counter_total(&rows, buffered_write::BYTES_COPIED_TOTAL), + Some(2048), + "the write helper must count the write size, not the call" + ); + assert_eq!(counter_total(&rows, "rustfs_zero_copy_memory_saved_bytes_total"), Some(1024)); + assert_eq!(counter_total(&rows, "rustfs_zero_copy_bytes_saved_total"), Some(4096)); + assert_eq!(histogram_samples(&rows, "rustfs_zero_copy_read_duration_ms"), vec![10.5]); + assert_eq!(histogram_samples(&rows, "rustfs_zero_copy_write_duration_ms"), vec![20.5]); + assert_eq!( + counter_total(&rows, "rustfs_bytes_pool_hits_total"), + Some(1), + "only the from_pool acquisition counts as a hit" + ); + assert_eq!( + counter_total(&rows, "rustfs_bytes_pool_misses_total"), + Some(1), + "only the non-pool acquisition counts as a miss" + ); + assert_eq!( + gauge_value(&rows, "rustfs_bytes_pool_hit_rate"), + Some(85.0), + "the hit rate is exported as a percentage" + ); } #[test] @@ -2685,20 +2813,6 @@ mod tests { set_get_stage_metrics_enabled(false); } - #[test] - fn test_record_zero_copy_write() { - record_zero_copy_write(1024, 10.5); - record_zero_copy_write_fallback("test"); - record_bytes_saved(1024); - } - - // S3 Operation Metrics Tests - #[test] - fn test_record_get_object() { - record_get_object(100.0, 1024 * 1024); - record_get_object(50.0, 2048); - } - #[test] fn test_record_get_object_stage_metrics() { record_get_object_stage_duration("s3_handler", "request_context", 0.001); @@ -2847,17 +2961,95 @@ mod tests { assert!(0.0003_f64.is_sign_positive()); } + /// Replaces five smoke tests (`test_record_get_object`, `test_record_put_object`, + /// `test_record_put_object_request_metrics`, `test_record_list_objects`, + /// `test_record_delete_object`) that called the S3 operation recorders and + /// asserted nothing (rustfs/backlog#1836). Besides pinning the metric names, + /// this pins the conditional emissions each helper owns: the zero-copy alias + /// counters fire only for an eligible PUT, the truncated/version counters only + /// for the truncated listing and the versioned delete. #[test] - fn test_record_put_object() { - record_put_object(200.0, 1024 * 1024, true); - record_put_object(100.0, 512, false); - } + fn s3_operation_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); - #[test] - fn test_record_put_object_request_metrics() { - record_put_object_request_start(3); - record_put_object_request_result("ok", 0.25); - record_put_object_request_result("error", 0.5); + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + set_get_stage_metrics_enabled(true); + record_get_object(100.0, 1024 * 1024); + record_get_object(50.0, 2048); + record_put_object(200.0, 1024 * 1024, true); + record_put_object(100.0, 512, false); + record_put_object_request_start(3); + record_put_object_request_result("ok", 0.25); + record_put_object_request_result("error", 0.5); + record_list_objects(50.0, 100, false); + record_list_objects(75.0, 1000, true); + record_delete_object(25.0, false); + record_delete_object(30.0, true); + set_get_stage_metrics_enabled(false); + set_metrics_enabled(false); + }); + + let rows = snapshotter.snapshot().into_vec(); + let names = emitted_names(&rows); + for expected in [ + "rustfs_s3_get_object_total", + "rustfs_s3_get_object_duration_ms", + "rustfs_s3_get_object_size_bytes", + "rustfs_s3_put_object_total", + "rustfs_s3_put_object_duration_ms", + "rustfs_s3_put_object_size_bytes", + "rustfs_s3_put_object_zero_copy_enabled_total", + "rustfs_s3_put_object_zero_copy_eligible_total", + "rustfs_io_put_object_requests_total", + "rustfs_io_put_object_concurrent_requests", + "rustfs_io_put_object_request_results_total", + "rustfs_io_put_object_request_duration_seconds", + "rustfs_s3_list_objects_total", + "rustfs_s3_list_objects_duration_ms", + "rustfs_s3_list_objects_count", + "rustfs_s3_list_objects_truncated_total", + "rustfs_s3_delete_object_total", + "rustfs_s3_delete_object_duration_ms", + "rustfs_s3_delete_object_version_total", + ] { + assert!(names.contains(expected), "{expected} must be emitted by its record helper"); + } + + assert_eq!(counter_total(&rows, "rustfs_s3_get_object_total"), Some(2)); + assert_eq!(counter_total(&rows, "rustfs_s3_put_object_total"), Some(2)); + assert_eq!( + counter_total(&rows, "rustfs_s3_put_object_zero_copy_eligible_total"), + Some(1), + "only the zero-copy eligible PUT increments the eligibility counter" + ); + assert_eq!( + counter_total(&rows, "rustfs_s3_put_object_zero_copy_enabled_total"), + Some(1), + "the historical alias must stay in step with the eligibility counter" + ); + assert_eq!( + counter_total(&rows, "rustfs_s3_list_objects_truncated_total"), + Some(1), + "only the truncated listing increments the truncation counter" + ); + assert_eq!( + counter_total(&rows, "rustfs_s3_delete_object_version_total"), + Some(1), + "only the versioned delete increments the version counter" + ); + assert_eq!( + histogram_samples(&rows, "rustfs_s3_list_objects_count"), + vec![100.0, 1000.0], + "the object count, not the duration, belongs in the count histogram" + ); + assert_eq!( + gauge_value(&rows, "rustfs_io_put_object_concurrent_requests"), + Some(3.0), + "the concurrency gauge must carry the reported in-flight request count" + ); } #[test] @@ -3175,74 +3367,171 @@ mod tests { set_metrics_enabled(false); } + /// Replaces six smoke tests (`test_record_io_strategy`, `test_record_permit_wait`, + /// `test_record_io_load_level`, `test_record_cache_size`, `test_record_bandwidth`, + /// `test_record_data_transfer`) that called the scheduler, cache and bandwidth + /// recorders and asserted nothing (rustfs/backlog#1836). The derived bandwidth + /// value and the `all` tier fan-out are now pinned, not just the names. #[test] - fn test_record_list_objects() { - record_list_objects(50.0, 100, false); - record_list_objects(75.0, 1000, true); + fn io_scheduler_and_bandwidth_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); + + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_io_strategy("nvme", "sequential", 256 * 1024, 5); + record_io_strategy("ssd", "random", 64 * 1024, 10); + record_permit_wait(5.0); + record_permit_wait(10.5); + record_io_load_level("low", 2); + record_io_load_level("high", 15); + record_cache_size("l1", 50 * 1024 * 1024, 1000); + record_bandwidth(100 * 1024 * 1024, "high"); + record_data_transfer(1024 * 1024, 100.0); + record_data_transfer(2048, 50.0); + set_metrics_enabled(false); + }); + + let rows = snapshotter.snapshot().into_vec(); + let names = emitted_names(&rows); + for expected in [ + "rustfs_io_strategy_total", + "rustfs_io_buffer_size_bytes", + "rustfs_io_concurrent_requests", + "rustfs_io_permit_wait_duration_ms", + "rustfs_io_load_level", + "rustfs_cache_size_bytes", + "rustfs_cache_entries", + "rustfs_bandwidth_current_bps", + "rustfs_bandwidth_observed_bps", + "rustfs_io_transfer_bytes_total", + "rustfs_io_transfer_duration_ms", + "rustfs_io_transfer_bandwidth_bps", + ] { + assert!(names.contains(expected), "{expected} must be emitted by its record helper"); + } + + assert_eq!(counter_total(&rows, "rustfs_io_strategy_total"), Some(2)); + assert_eq!(counter_total(&rows, "rustfs_io_load_level"), Some(2)); + assert_eq!( + gauge_value(&rows, "rustfs_io_concurrent_requests"), + Some(15.0), + "the shared concurrency gauge must hold the last reported value" + ); + assert_eq!(histogram_samples(&rows, "rustfs_io_permit_wait_duration_ms"), vec![5.0, 10.5]); + assert_eq!(gauge_value(&rows, "rustfs_cache_entries"), Some(1000.0)); + assert_eq!(gauge_value(&rows, "rustfs_cache_size_bytes"), Some((50 * 1024 * 1024) as f64)); + assert_eq!( + counter_total(&rows, "rustfs_io_transfer_bytes_total"), + Some(1024 * 1024 + 2048), + "transferred bytes must accumulate across calls" + ); + assert_eq!( + histogram_samples(&rows, "rustfs_io_transfer_bandwidth_bps"), + vec![40960.0, 10_485_760.0], + "bandwidth must be derived as bytes * 1000 / duration_ms" + ); + + let mut bandwidth_by_tier: Vec<(&str, f64)> = rows + .iter() + .filter(|(composite, _, _, _)| composite.key().name() == "rustfs_bandwidth_current_bps") + .map(|(composite, _, _, value)| { + let tier = composite + .key() + .labels() + .find(|label| label.key() == "tier") + .map(|label| label.value()) + .expect("bandwidth gauges carry a tier label"); + match value { + DebugValue::Gauge(value) => (tier, value.0), + other => panic!("rustfs_bandwidth_current_bps holds {other:?}"), + } + }) + .collect(); + bandwidth_by_tier.sort_by(|left, right| left.0.cmp(right.0)); + assert_eq!( + bandwidth_by_tier, + vec![("all", 104_857_600.0), ("high", 104_857_600.0)], + "record_bandwidth must publish both the aggregate `all` series and the caller tier" + ); } + /// Replaces five smoke tests (`test_record_memory_usage`, + /// `test_record_process_memory_split`, `test_record_cgroup_memory_split`, + /// `test_record_cpu_usage`, `test_record_disk_io`) that called the system + /// resource recorders and asserted nothing (rustfs/backlog#1836). The gauge + /// values pin the argument order and the usage-percent derivation, which name + /// checks alone cannot catch. #[test] - fn test_record_delete_object() { - record_delete_object(25.0, false); - record_delete_object(30.0, true); + fn system_resource_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); + + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_memory_usage(1024 * 1024 * 1024, 4 * 1024 * 1024 * 1024); + record_process_memory_split(1024, 2048); + record_cgroup_memory_split(Some(1), Some(2), Some(3), Some(4), Some(5), Some(6)); + record_cpu_usage(25.5); + record_disk_io(1024 * 1024, 2048, 100, 50); + record_disk_io(2048, 4096, 200, 100); + set_metrics_enabled(false); + }); + + let rows = snapshotter.snapshot().into_vec(); + assert_eq!(gauge_value(&rows, "rustfs_memory_used_bytes"), Some((1024 * 1024 * 1024) as f64)); + assert_eq!(gauge_value(&rows, "rustfs_memory_total_bytes"), Some((4u64 * 1024 * 1024 * 1024) as f64)); + assert_eq!( + gauge_value(&rows, "rustfs_memory_usage_percent"), + Some(25.0), + "usage percent must be used/total * 100" + ); + assert_eq!(gauge_value(&rows, "rustfs_memory_process_resident_bytes"), Some(1024.0)); + assert_eq!( + gauge_value(&rows, "rustfs_memory_process_virtual_bytes"), + Some(2048.0), + "resident and virtual bytes must not be swapped" + ); + for (name, expected) in [ + ("rustfs_memory_cgroup_current_bytes", 1.0), + ("rustfs_memory_cgroup_limit_bytes", 2.0), + ("rustfs_memory_cgroup_anon_bytes", 3.0), + ("rustfs_memory_cgroup_file_bytes", 4.0), + ("rustfs_memory_cgroup_active_file_bytes", 5.0), + ("rustfs_memory_cgroup_inactive_file_bytes", 6.0), + ] { + assert_eq!(gauge_value(&rows, name), Some(expected), "{name} must receive its own argument"); + } + assert_eq!(gauge_value(&rows, "rustfs_cpu_usage_percent"), Some(25.5)); + assert_eq!(counter_total(&rows, "rustfs_disk_read_bytes_total"), Some(1024 * 1024 + 2048)); + assert_eq!(counter_total(&rows, "rustfs_disk_write_bytes_total"), Some(2048 + 4096)); + assert_eq!(counter_total(&rows, "rustfs_disk_read_ops_total"), Some(300)); + assert_eq!( + counter_total(&rows, "rustfs_disk_write_ops_total"), + Some(150), + "byte and op counters must not be crossed" + ); } - // I/O Scheduler Metrics Tests + /// Boundary companion of the `Some(..)` case above: an absent cgroup field must + /// emit no gauge at all. Publishing `0` for a field the kernel does not expose + /// would read as a real measurement (rustfs/backlog#1836). #[test] - fn test_record_io_strategy() { - record_io_strategy("nvme", "sequential", 256 * 1024, 5); - record_io_strategy("ssd", "random", 64 * 1024, 10); - } + fn cgroup_memory_split_skips_absent_fields() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); - #[test] - fn test_record_permit_wait() { - record_permit_wait(5.0); - record_permit_wait(10.5); - } + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_cgroup_memory_split(None, None, None, None, None, None); + set_metrics_enabled(false); + }); - #[test] - fn test_record_io_load_level() { - record_io_load_level("low", 2); - record_io_load_level("medium", 5); - record_io_load_level("high", 15); - } - - #[test] - fn test_record_cache_size() { - record_cache_size("l1", 50 * 1024 * 1024, 1000); - record_cache_size("l2", 200 * 1024 * 1024, 5000); - } - - // Bandwidth Metrics Tests - #[test] - fn test_record_bandwidth() { - record_bandwidth(100 * 1024 * 1024, "high"); - record_bandwidth(50 * 1024 * 1024, "medium"); - } - - #[test] - fn test_record_data_transfer() { - record_data_transfer(1024 * 1024, 100.0); - record_data_transfer(2048, 50.0); - } - - // System Resource Metrics Tests - #[test] - fn test_record_memory_usage() { - record_memory_usage(1024 * 1024 * 1024, 4 * 1024 * 1024 * 1024); - record_memory_usage(2 * 1024 * 1024 * 1024, 8 * 1024 * 1024 * 1024); - } - - #[test] - fn test_record_process_memory_split() { - record_process_memory_split(1024, 2048); - record_process_memory_split(4096, 8192); - } - - #[test] - fn test_record_cgroup_memory_split() { - record_cgroup_memory_split(Some(1), Some(2), Some(3), Some(4), Some(5), Some(6)); - record_cgroup_memory_split(None, None, None, None, None, None); + let rows = snapshotter.snapshot().into_vec(); + assert!(rows.is_empty(), "absent cgroup fields must emit nothing, got {:?}", emitted_names(&rows)); } #[test] @@ -3460,36 +3749,75 @@ mod tests { assert_eq!(current_get_object_buffered_bytes(), 0); } + /// Replaces three smoke tests (`test_record_error`, `test_record_timeout`, + /// `test_record_retry`) that called the failure recorders and asserted nothing + /// (rustfs/backlog#1836). The histogram samples pin that the timeout duration + /// and the retry attempt number reach their histogram rather than being folded + /// into the counters. #[test] - fn test_record_cpu_usage() { - record_cpu_usage(25.5); - record_cpu_usage(50.0); - record_cpu_usage(75.5); - } + fn failure_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); - #[test] - fn test_record_disk_io() { - record_disk_io(1024 * 1024, 2048, 100, 50); - record_disk_io(2048, 4096, 200, 100); - } + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_error("get_object", "timeout"); + record_error("put_object", "disk_error"); + record_timeout("get_object", 5000.0); + record_timeout("list_objects", 10000.0); + record_retry("get_object", 1); + record_retry("put_object", 2); + set_metrics_enabled(false); + }); - // Error and Timeout Metrics Tests - #[test] - fn test_record_error() { - record_error("get_object", "timeout"); - record_error("put_object", "disk_error"); - } + let rows = snapshotter.snapshot().into_vec(); + let names = emitted_names(&rows); + for expected in [ + "rustfs_errors_total", + "rustfs_timeouts_total", + "rustfs_timeouts_duration_ms", + "rustfs_retries_total", + "rustfs_retries_attempt", + ] { + assert!(names.contains(expected), "{expected} must be emitted by its record helper"); + } - #[test] - fn test_record_timeout() { - record_timeout("get_object", 5000.0); - record_timeout("list_objects", 10000.0); - } + assert_eq!( + counter_total(&rows, "rustfs_errors_total"), + Some(2), + "each error is counted once under its own operation/type labels" + ); + assert_eq!(counter_total(&rows, "rustfs_timeouts_total"), Some(2)); + assert_eq!(counter_total(&rows, "rustfs_retries_total"), Some(2)); + assert_eq!(histogram_samples(&rows, "rustfs_timeouts_duration_ms"), vec![5000.0, 10000.0]); + assert_eq!( + histogram_samples(&rows, "rustfs_retries_attempt"), + vec![1.0, 2.0], + "the attempt number belongs in the histogram, not the retry counter" + ); - #[test] - fn test_record_retry() { - record_retry("get_object", 1); - record_retry("put_object", 2); + let mut error_labels: Vec<(&str, &str)> = rows + .iter() + .filter(|(composite, _, _, _)| composite.key().name() == "rustfs_errors_total") + .map(|(composite, _, _, _)| { + let label = |key: &str| { + composite + .key() + .labels() + .find(|label| label.key() == key) + .map(|label| label.value()) + .expect("error counters carry operation and type labels") + }; + (label("operation"), label("type")) + }) + .collect(); + error_labels.sort(); + assert_eq!( + error_labels, + vec![("get_object", "timeout"), ("put_object", "disk_error")], + "operation and error type must not be swapped" + ); } } @@ -3633,42 +3961,116 @@ pub fn update_zero_copy_performance_metrics(copy_count: u32, throughput_mbps: f6 #[cfg(test)] mod zero_copy_tests { use super::*; + use crate::tests::{METRICS_FLAG_LOCK, counter_total, emitted_names, gauge_value, histogram_samples}; + use metrics_util::debugging::DebuggingRecorder; + /// Replaces six smoke tests (`test_record_zero_copy_buffer_operation`, + /// `test_record_memory_copy`, `test_record_shared_ref_operation`, + /// `test_record_bufreader_optimization`, `test_record_direct_io_operation`, + /// `test_update_zero_copy_performance_metrics`) whose own comment admitted they + /// only checked that the helpers compile and run (rustfs/backlog#1836). The same + /// calls now run against a local recorder, and the assertions pin the counter + /// split (operations vs bytes, copies vs copied bytes), the success/fallback + /// label mapping, and the three same-typed performance gauges. #[test] - fn test_record_zero_copy_buffer_operation() { - // This test verifies the function compiles and runs - // Actual metric verification requires a metrics recorder - record_zero_copy_buffer_operation("read", 1024); - record_zero_copy_buffer_operation("write", 2048); - } + fn zero_copy_helpers_emit_their_metrics() { + let _guard = METRICS_FLAG_LOCK.lock().unwrap_or_else(|e| e.into_inner()); + let recorder = DebuggingRecorder::new(); + let snapshotter = recorder.snapshotter(); - #[test] - fn test_record_memory_copy() { - record_memory_copy(1, 1024); - record_memory_copy(2, 2048); - } + metrics::with_local_recorder(&recorder, || { + set_metrics_enabled(true); + record_zero_copy_buffer_operation("read", 1024); + record_zero_copy_buffer_operation("write", 2048); + record_memory_copy(1, 1024); + record_memory_copy(2, 2048); + record_shared_ref_operation("create"); + record_shared_ref_operation("share"); + record_bufreader_optimization(1, 8192); + record_bufreader_optimization(2, 65536); + record_direct_io_operation("read", 4096, true); + record_direct_io_operation("write", 8192, false); + update_zero_copy_performance_metrics(2, 150.5, 1024 * 1024); + set_metrics_enabled(false); + }); - #[test] - fn test_record_shared_ref_operation() { - record_shared_ref_operation("create"); - record_shared_ref_operation("share"); - } + let rows = snapshotter.snapshot().into_vec(); + let names = emitted_names(&rows); + for expected in [ + zero_copy::BUFFER_OPERATIONS_TOTAL, + zero_copy::BUFFER_BYTES_TOTAL, + zero_copy::MEMORY_COPY_TOTAL, + zero_copy::MEMORY_COPY_BYTES_TOTAL, + "rustfs_memory_copy_size_bytes", + zero_copy::SHARED_REF_OPERATIONS_TOTAL, + zero_copy::BUFREADER_LAYERS_ELIMINATED_TOTAL, + zero_copy::BUFREADER_BUFFER_SIZE_BYTES, + zero_copy::DIRECT_IO_OPERATIONS_TOTAL, + zero_copy::DIRECT_IO_BYTES_TOTAL, + aligned_pread::OPERATIONS_TOTAL, + aligned_pread::BYTES_TOTAL, + zero_copy::AVG_COPY_COUNT, + zero_copy::THROUGHPUT_MBPS, + zero_copy::MEMORY_SAVED_BYTES, + ] { + assert!(names.contains(expected), "{expected} must be emitted by its record helper"); + } - #[test] - fn test_record_bufreader_optimization() { - record_bufreader_optimization(1, 8192); - record_bufreader_optimization(2, 65536); - } + assert_eq!(counter_total(&rows, zero_copy::BUFFER_OPERATIONS_TOTAL), Some(2)); + assert_eq!( + counter_total(&rows, zero_copy::BUFFER_BYTES_TOTAL), + Some(3072), + "buffer bytes must accumulate the sizes, not the call count" + ); + assert_eq!( + counter_total(&rows, zero_copy::MEMORY_COPY_TOTAL), + Some(3), + "the copy counter takes the copy count argument" + ); + assert_eq!( + counter_total(&rows, zero_copy::MEMORY_COPY_BYTES_TOTAL), + Some(3072), + "the copied-bytes counter takes the size argument" + ); + assert_eq!(histogram_samples(&rows, "rustfs_memory_copy_size_bytes"), vec![1024.0, 2048.0]); + assert_eq!(counter_total(&rows, zero_copy::SHARED_REF_OPERATIONS_TOTAL), Some(2)); + assert_eq!(counter_total(&rows, zero_copy::BUFREADER_LAYERS_ELIMINATED_TOTAL), Some(3)); + assert_eq!(histogram_samples(&rows, zero_copy::BUFREADER_BUFFER_SIZE_BYTES), vec![8192.0, 65536.0]); + assert_eq!(counter_total(&rows, zero_copy::DIRECT_IO_BYTES_TOTAL), Some(12288)); + assert_eq!( + counter_total(&rows, aligned_pread::BYTES_TOTAL), + Some(12288), + "the aligned-pread series must mirror the direct-IO series" + ); + assert_eq!(gauge_value(&rows, zero_copy::AVG_COPY_COUNT), Some(2.0)); + assert_eq!(gauge_value(&rows, zero_copy::THROUGHPUT_MBPS), Some(150.5)); + assert_eq!( + gauge_value(&rows, zero_copy::MEMORY_SAVED_BYTES), + Some((1024 * 1024) as f64), + "the three performance gauges must not be filled from each other's argument" + ); - #[test] - fn test_record_direct_io_operation() { - record_direct_io_operation("read", 4096, true); - record_direct_io_operation("write", 8192, false); - } - - #[test] - fn test_update_zero_copy_performance_metrics() { - update_zero_copy_performance_metrics(2, 150.5, 1024 * 1024); + let mut direct_io_labels: Vec<(&str, &str)> = rows + .iter() + .filter(|(composite, _, _, _)| composite.key().name() == zero_copy::DIRECT_IO_OPERATIONS_TOTAL) + .map(|(composite, _, _, _)| { + let label = |key: &str| { + composite + .key() + .labels() + .find(|label| label.key() == key) + .map(|label| label.value()) + .expect("direct-IO counters carry operation and status labels") + }; + (label("operation"), label("status")) + }) + .collect(); + direct_io_labels.sort(); + assert_eq!( + direct_io_labels, + vec![("read", "success"), ("write", "fallback")], + "the success flag must map to the success/fallback status label" + ); } #[test] diff --git a/scripts/find_assertless_tests.py b/scripts/find_assertless_tests.py index 90e52a21f..dc73447de 100755 --- a/scripts/find_assertless_tests.py +++ b/scripts/find_assertless_tests.py @@ -28,8 +28,15 @@ called helper. Known false-positive classes are excluded up front: parameters; the assert lives in the shared body — still scanned, but a body that asserts is not flagged anyway; the exclusion covers wrappers that only delegate to a suite runner). -- Functions whose body calls a helper with `assert`, `verify`, `check`, - `expect`, `run_` or `_case` in its name (suite-delegation pattern). +- Functions whose body calls a helper *named* like a shared check or suite + runner: an `assert_`/`verify_`/`check_`/`expect_`/`ensure_`/`run_` prefix, + or a `_case`/`_cases`/`_harness`/`_roundtrip` suffix. The name must carry + the token as its own leading or trailing segment — matching it anywhere + inside the identifier hid whole test bodies behind an unrelated domain + call such as `record_get_object_bitrot_verify_duration(..)`. +- Functions whose body only defines an unused inner `fn _name(..)`: that is + the compile-time shape check (exhaustive match, signature pin), where the + type system is the assertion. Usage: scripts/find_assertless_tests.py [path ...] # default: crates rustfs/src @@ -45,7 +52,11 @@ VERIFY_SIGNALS = re.compile( r"assert!|assert_eq!|assert_ne!|debug_assert|panic!\(|\.expect\(|\.unwrap\(|" r"unreachable!|matches!\(|insta::|proptest!|\.await\?|\)\?|\?;|should_panic" ) -DELEGATION = re.compile(r"\b[a-z0-9_]*(?:assert|verify|check|expect|run_case|_case|harness|round_trip|roundtrip)[a-z0-9_]*\s*\(") +DELEGATION = re.compile( + r"\b(?:assert|verify|check|expect|ensure|run)_[a-z0-9_]*\s*\(|" + r"\b[a-z0-9_]+_(?:case|cases|harness|roundtrip|round_trip)\s*\(" +) +COMPILE_TIME_CHECK = re.compile(r"\bfn\s+_[a-zA-Z0-9_]*\s*(?:<[^>]*>)?\s*\(") TEST_ATTR = re.compile(r"#\[(?:tokio::)?test[\](]") TEST_CASE_ATTR = re.compile(r"#\[test_case") FN_LINE = re.compile(r"^\s*(?:pub\s+)?(?:async\s+)?fn\s+([a-zA-Z0-9_]+)") @@ -94,7 +105,7 @@ def scan_file(path: Path): break k += 1 text = "\n".join(body) - if not VERIFY_SIGNALS.search(text) and not DELEGATION.search(text): + if not VERIFY_SIGNALS.search(text) and not DELEGATION.search(text) and not COMPILE_TIME_CHECK.search(text): print(f"{path}:{j + 1}: {name}") i = k + 1 From 1cf0f7af1583caeafea278c082688814017ba6af Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=94=90=E5=B0=8F=E9=B8=AD?= Date: Tue, 18 Aug 2026 21:45:38 +0800 Subject: [PATCH 19/24] feat(replication): split oversized hot-path functions, proxy unreplicated reads, and fail SSE-C passthrough closed (#6170) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * refactor(replication): split four oversized hot-path functions into focused helpers Pure-move decomposition of the four oversized functions flagged by the replication compatibility review (P1-18), unblocking migration milestone M2 which requires resyncer moves to stay mechanical: - resync_bucket (522 lines -> 61-line step sequence): leader lock, target resolution, walk/collector/worker spawning, and dispatch loop extracted into focused helpers; pure decision helpers (DTO builders, HEAD-result classification) separated from IO orchestration. - replicate_all (411 lines -> 113-line main body): initial target-info seeding, read/stat option builders, skip-path notes, target HEAD action resolution, and the multipart/single-put payload transport extracted as private free functions. - start_mrf_processor (306 lines -> 46-line spawn body): recovery guard, ledger load, per-entry replay (delete/object/metadata), and retained entry resolution extracted; retry bookkeeping semantics preserved exactly (inner continue-paths push inside helpers, outer Missed push stays in the loop). - apply_iam_item (255 lines -> match dispatch skeleton): one helper per IAM item type. No behavior change: log texts, error paths, event emissions, and metric counts are byte-identical; existing tests unchanged and green (238 ecstore replication/mrf/resync + 232 rustfs site-replication). * feat(replication): proxy GET/HEAD/Tagging for unreplicated objects to replication targets (#6172) * feat(replication): proxy GET/HEAD/Tagging for unreplicated objects to replication targets Implements the MinIO active-active read-proxy protocol (P1-5 of the replication compatibility review): when a GET/HEAD/GetObjectTagging/ PutObjectTagging/DeleteObjectTagging request fails locally with not-found and the bucket has replication targets, the request is proxied to the targets in rule order, mirroring bucket-replication.go proxyGetToReplicationTarget/proxyHeadToRepTarget/proxyTaggingToRepTarget. Protocol surface: - Anti-loop: inbound {x-rustfs-,x-minio-}source-proxy-request is parsed into ObjectOptions (proxy_request + proxy_header_set, matching MinIO ProxyRequest/ProxyHeaderSet); a request carrying the marker with ANY value is never re-proxied. Outbound client proxy calls send the marker as "true"; replication worker convergence HEADs send it as "false" so a peer's proxy layer cannot answer a convergence check by proxying back to the source (which would fake Completed without a PUT). - Target selection: new replication_proxy.rs get_proxy_targets — empty when the marker is set, versioning is suspended, or no replication config; otherwise filter_target_arns -> TargetClient lookup, skipping targets with proxying disabled. - TargetClient gains head_object_for_proxy/get_object (streaming) and the three tagging calls. Proxy calls never send the replication-check SSE-C exemption header; customer SSE-C keys are forwarded verbatim so the target performs real decryption. Conditional (If-*) headers are not forwarded (MinIO parity); Range and part_number are, with parts_count/tag_count/storage_class/expiration passed through. - Metrics: proxy counters now count only real client proxy traffic, MinIO-aligned (one total per proxied request, one failed when no target served it). The previous misattributed counters — replication worker HEAD/PUT (#2672) and local tagging operations (#2682) — are removed; ReplProxyMetric now maps the tagging counters instead of dropping them. e2e (fake_s3_target extended with tagging + header journaling): proxied GET body + outbound header contract (marker present, no replication-check, SSE-C passthrough), HEAD, anti-loop 404 with zero outbound requests, GetObjectTagging, and metric mapping unit tests. Rolling note: proxying only activates for buckets with replication targets; requests carrying the marker keep pre-upgrade behavior. Refs rustfs/backlog#1675 (P1-5) * fix(replication): fail SSE-C passthrough closed on targets that drop transport headers (#6178) SSE-C ciphertext passthrough replicates via X-Rustfs-Replication-* transport headers. A MinIO/generic-S3 target silently discards them, storing bare ciphertext with no decryption material — yet the PUT succeeded, so the object reported COMPLETED with a silently unreadable replica (backlog#1675 N2). Fail-closed design: - SsecPassthroughCapability {Unknown, Supported, Unsupported} cached in BucketTargetSys per target ARN with a recording timestamp. Entries reset whenever the target is rebuilt, edited, or removed (arn_remotes_map lifecycle) and expire after SSEC_PASSTHROUGH_CAPABILITY_TTL (10 minutes): an expired verdict in either direction is re-earned through the audit, so an Unsupported target recovers automatically after an upgrade (at most one wasted PUT+HEAD audit per bad target per TTL window) and a Supported verdict cannot outlive a backend swapped behind the same endpoint. - Replication worker (replicate_object and replicate_all): fresh Unsupported targets never receive the PUT — the attempt fails immediately into the normal MRF retry channel with a "run ?replication-check to re-probe" hint. Unknown or expired verdicts are audited: after the PUT the worker HEADs the replica back through the replication-check channel (source version id mapped through resolve_read_api_version_id, so null-version objects audit correctly) and requires SSE-C evidence (the echoed customer-algorithm header); missing evidence records Unsupported and fails the attempt. Convergence HEADs are audited the same way, so a broken ciphertext replica from an earlier attempt can never launder itself into COMPLETED via an ETag match. The gate/evidence policy is pure (replication_target_boundary, staleness folded in as an input) for the M2 worker migration. - replication-check grows an SsecPassthrough probe phase: a probe PUT carrying the live transport-header shape, HEAD-back for evidence, and a machine-readable Code BucketRemoteSsecPassthroughUnsupported on failure. The probe verdict is synced into the runtime capability cache. Unlike VersionFidelity, a failed SsecPassthrough phase does NOT fail the target overall — it is a capability limit, not a broken replication contract, and a plaintext-only deployment against such a target must not turn red. - fake_s3_target: default mode now models a RustFS target (stores the transport headers, echoes SSE-C evidence); the new drop_unlisted_replication_headers mode models MinIO. The journal records whether a request carried transport headers. Receiver-echo verification: the replication-check HEAD exemption only skips SSE-C key validation; the response has always built sse-customer-algorithm from stored metadata (rustfs/src/app/object_usecase.rs), so no receiver change was needed — pinned end to end by the replication-check e2e against a real RustFS target. Rolling-upgrade constraint: RustFS targets older than the replication-check HEAD exemption (#5898) answer the audit HEAD without SSE-C evidence (or fail it outright), so SSE-C replication to such targets reports FAILED. This is deliberate — FAILED-and-retryable beats a silently undecryptable replica — and self-heals: once the target is upgraded, the next TTL expiry (or a manual ?replication-check re-probe) re-audits and records Supported. Plaintext and managed-SSE replication are unaffected. The capability cache is per-node; each node audits independently. Known limitations: - The audit judges evidence from the echoed customer-algorithm header only. A hypothetical target that preserves that one header while dropping other transport headers (partial-drop) would pass the audit; no known target behaves this way — observed targets drop the whole unknown-header family. - A mixed-version target cluster can flap the verdict between audits routed to different target nodes until the rollout completes; the TTL bounds how long each stale verdict persists. New e2e (backlog#1675 C1 + N2, red-first): fail-closed against a header-dropping fake (FAILED + no second PUT via the capability cache, journal-asserted; red run showed the old COMPLETED), replication-check reports the SsecPassthrough phase Code while the target stays OK overall, SSE-C heal convergence after a real target outage, and SSE-C existing-object resync landing a REPLICA readable with the customer key. TTL expiry in both directions is pinned at the cache and gate seams. * refactor(replication): move resyncer pure decision logic into rustfs-replication (M2) (#6180) * refactor(replication): move resyncer pure decision logic into rustfs-replication (M2) Pure-move milestone M2 of the ECStore replication split (backlog#1675 P1-17): relocate the resyncer's IO-free decision helpers, with their unit tests, into the crates they already belong to by type ownership. No behavior change. Moved into crates/replication: - resync.rs: resync_status_duration - delete.rs: resync_existing_delete_replication_info, replicate_delete_outcome, target_delete_version_id, delete_marker_purge_version_id, delete_marker_purge_mrf_entry - object.rs: version_identity_drifted, is_replication_target_offline_error, SsecPassthroughCapability, SsecPassthroughGate, ssec_passthrough_gate, ssec_passthrough_evidence_present (param-demoted to the echoed customer-algorithm string; ECStore keeps the HeadObjectOutput adapter) - filemeta.rs: NULL_VERSION_ID wire literal (crate-owned copy per the filemeta-independence contract) ECStore rewiring (Rule #14: imports stay in *_boundary.rs): - resync/object-decision/target boundaries re-export the moved symbols; resyncer call sites are unchanged - bucket_target_sys keeps only the verdict cache + TTL and re-exports the capability enum so existing consumer paths keep compiling Not moved (signatures carry ECStore or aws-sdk types): verify_resync_head_result, resync_target_error_detail, the SdkError classifiers, the replicate_all_* option/info builders, and the env-coupled bounded_resync_max_jobs admission clamp. README milestone table updated. * chore(replication): retire the datatypes.rs relay early README sanctions retiring datatypes.rs ahead of M4. The module was a pure relay (resync boundary -> datatypes -> mod.rs facade) with no external consumer importing it directly, so the facade now re-exports ResyncStatusType from replication_resync_boundary and the relay file is deleted. Consumers stay behind the ECStore facade, keeping Migration Rule #15 intact — the original retirement wording ("consumers import through rustfs-replication directly") conflicted with that rule and is corrected in the README. * chore(arch): extend migration guards to the M2-moved decision contracts The adversarial review of the M2 move found the per-symbol ratchet in check_architecture_migration_rules.sh was not extended for the moved symbols, leaving them free to be redefined in ECStore or imported past their boundary without CI noticing: - resync definition pin + boundary fences gain resync_status_duration; - the object-decision boundary fences gain the five delete-family helpers (delete_marker_purge_mrf_entry, delete_marker_purge_version_id, replicate_delete_outcome, resync_existing_delete_replication_info, target_delete_version_id); - the target-boundary fence gains the SSE-C gate family, the offline classifier, and version_identity_drifted; - a new definition pin rejects ECStore redefinitions of the M2-moved fns/enums (ssec_passthrough_evidence_present deliberately excluded: ECStore keeps a thin HeadObjectOutput adapter under that name). Mutation-verified: a probe fn ssec_passthrough_gate under crates/ecstore/src/bucket/replication trips the new pin. Also anchors the intentionally-duplicated NULL_VERSION_ID wire literal from the filemeta side and tightens the M2 README note on bounded_resync_max_jobs. --- crates/e2e_test/src/fake_s3_target/README.md | 4 +- crates/e2e_test/src/fake_s3_target/mod.rs | 289 ++- .../src/replication_extension_test.rs | 722 +++++- crates/ecstore/src/api/mod.rs | 15 +- .../ecstore/src/bucket/bucket_target_sys.rs | 310 ++- .../ecstore/src/bucket/replication/README.md | 15 +- .../src/bucket/replication/datatypes.rs | 15 - crates/ecstore/src/bucket/replication/mod.rs | 6 +- .../replication_filemeta_boundary.rs | 7 +- .../replication_object_decision_boundary.rs | 11 +- .../bucket/replication/replication_pool.rs | 636 ++--- .../bucket/replication/replication_proxy.rs | 150 ++ .../replication_resync_boundary.rs | 9 +- .../replication/replication_resyncer.rs | 2044 +++++++++-------- .../bucket/replication/replication_state.rs | 25 + .../replication_target_boundary.rs | 47 +- crates/ecstore/src/object_api/types.rs | 14 + crates/filemeta/src/fileinfo.rs | 3 + crates/replication/src/delete.rs | 206 +- crates/replication/src/filemeta.rs | 5 + crates/replication/src/lib.rs | 14 +- crates/replication/src/object.rs | 213 +- crates/replication/src/resync.rs | 41 + rustfs/src/admin/handlers/site_replication.rs | 504 ++-- rustfs/src/admin/router.rs | 270 ++- rustfs/src/admin/storage_api.rs | 1 + rustfs/src/app/object_usecase.rs | 251 +- rustfs/src/app/storage_api.rs | 18 + rustfs/src/storage/ecfs.rs | 267 ++- rustfs/src/storage/mod.rs | 36 +- rustfs/src/storage/options.rs | 96 +- rustfs/src/storage/storage_api.rs | 26 +- scripts/check_architecture_migration_rules.sh | 36 +- 33 files changed, 4694 insertions(+), 1612 deletions(-) delete mode 100644 crates/ecstore/src/bucket/replication/datatypes.rs create mode 100644 crates/ecstore/src/bucket/replication/replication_proxy.rs diff --git a/crates/e2e_test/src/fake_s3_target/README.md b/crates/e2e_test/src/fake_s3_target/README.md index 78597195c..7f1feae0b 100644 --- a/crates/e2e_test/src/fake_s3_target/README.md +++ b/crates/e2e_test/src/fake_s3_target/README.md @@ -4,8 +4,8 @@ This module is the shared failure-injection boundary for replication end-to-end `FakeS3Target::start()` creates the listener. Add target buckets with `create_bucket`, point a RustFS remote target at `address()`, use `FAKE_ACCESS_KEY` / `FAKE_SECRET_KEY`, then enqueue per-operation faults with `inject`. Faults for one operation are consumed in FIFO order and do not consume faults queued for another operation. A fault is consumed only after `s3s` verifies the full request signature, so anonymous, other-access-key, and bad-signature traffic cannot disturb a script. -Supported data operations are HeadBucket, GetBucketVersioning, PUT/GET/HEAD/DELETE Object, and create/upload/complete/abort multipart upload. `create_bucket` models general-purpose buckets in S3's shared global namespace; account-regional namespace buckets and their `-an` names are intentionally out of scope. Buckets are versioned: PUT creates a version, DELETE without `versionId` creates a delete marker, and DELETE with `versionId` removes exactly that version. Internal source version IDs must be UUIDs and are stored canonically. Source mtime is honored only for source-replication PUT/DELETE requests; absent or invalid values use receipt time, matching RustFS, while multipart completion always uses receipt time. Replicated versions are ordered newest-first by source mtime so late older versions and delete markers do not become current. Equal mtimes prefer objects over delete markers, then canonical UUID order; RustFS's internal FileMeta signature tie-break is intentionally out of scope because it is not part of the target S3 protocol. Multipart part numbers follow S3's `1..=10000` range, and every completed part except the final part must be at least 5 MiB. +Supported data operations are HeadBucket, GetBucketVersioning, PUT/GET/HEAD/DELETE Object, Get/Put/Delete ObjectTagging (tags live per version; Put replaces the whole set, Delete clears it), and create/upload/complete/abort multipart upload. `create_bucket` models general-purpose buckets in S3's shared global namespace; account-regional namespace buckets and their `-an` names are intentionally out of scope. Buckets are versioned: PUT creates a version, DELETE without `versionId` creates a delete marker, and DELETE with `versionId` removes exactly that version. Internal source version IDs must be UUIDs and are stored canonically. Source mtime is honored only for source-replication PUT/DELETE requests; absent or invalid values use receipt time, matching RustFS, while multipart completion always uses receipt time. Replicated versions are ordered newest-first by source mtime so late older versions and delete markers do not become current. Equal mtimes prefer objects over delete markers, then canonical UUID order; RustFS's internal FileMeta signature tie-break is intentionally out of scope because it is not part of the target S3 protocol. Multipart part numbers follow S3's `1..=10000` range, and every completed part except the final part must be at least 5 MiB. -Fault actions cover HTTP 401/403/503 responses, pre-dispatch delay, connection abort when a logical request-body threshold is reached, streaming slow drain, and a deliberately wrong response ETag (including multipart-complete XML). `requests()` returns the ordered, credential-free request journal for assertions. +Fault actions cover HTTP 401/403/503 responses, pre-dispatch delay, connection abort when a logical request-body threshold is reached, streaming slow drain, and a deliberately wrong response ETag (including multipart-complete XML). `requests()` returns the ordered, credential-free request journal for assertions. Each record also journals a `ProxyHeaderSnapshot` — the read-proxy anti-loop marker (`x-{rustfs,minio}-source-proxy-request`), the replication-check exemption header, and the client SSE-C header family (algorithm and key-MD5 values; for the key itself only its presence) — so proxy tests can pin the exact wire contract. The listener is loopback-only. It admits at most 64 active connections and two concurrently buffered request bodies; authenticated multipart-complete XML collection and assembly take both body permits. Keep-alive is disabled, request-header reads are bounded to 30 seconds, a parsed request is bounded to 65 seconds, and the complete connection lifetime is bounded to 100 seconds. It retains at most 256 buckets, 4,096 journal entries, 4,096 scripted faults, 4,096 object versions, 256 multipart uploads, and 10,000 multipart parts. Retained identifiers are capped at 1 KiB, user metadata at 2 KiB, and content type at 1 KiB. A PUT or uploaded part is capped at 64 MiB; a completed multipart object and all stored object/part data are capped at 128 MiB. Body drain, body-permit waits, delay, and slow-drain execution are bounded to 30 seconds; each slow-drain slice delay must be below that bound. diff --git a/crates/e2e_test/src/fake_s3_target/mod.rs b/crates/e2e_test/src/fake_s3_target/mod.rs index c8ddcecf3..aad643043 100644 --- a/crates/e2e_test/src/fake_s3_target/mod.rs +++ b/crates/e2e_test/src/fake_s3_target/mod.rs @@ -30,10 +30,12 @@ use s3s::access::{S3Access, S3AccessContext}; use s3s::auth::SimpleAuth; use s3s::dto::{ AbortMultipartUploadInput, AbortMultipartUploadOutput, CompleteMultipartUploadInput, CompleteMultipartUploadOutput, - CreateMultipartUploadInput, CreateMultipartUploadOutput, DeleteMarkerEntry, DeleteObjectInput, DeleteObjectOutput, ETag, - GetBucketVersioningInput, GetBucketVersioningOutput, GetObjectInput, GetObjectOutput, HeadBucketInput, HeadBucketOutput, + CreateMultipartUploadInput, CreateMultipartUploadOutput, DeleteMarkerEntry, DeleteObjectInput, DeleteObjectOutput, + DeleteObjectTaggingInput, DeleteObjectTaggingOutput, ETag, GetBucketVersioningInput, GetBucketVersioningOutput, + GetObjectInput, GetObjectOutput, GetObjectTaggingInput, GetObjectTaggingOutput, HeadBucketInput, HeadBucketOutput, HeadObjectInput, HeadObjectOutput, ListObjectVersionsInput, ListObjectVersionsOutput, ObjectVersionId, PutObjectInput, - PutObjectOutput, StreamingBlob, Timestamp, TimestampFormat, UploadPartInput, UploadPartOutput, + PutObjectOutput, PutObjectTaggingInput, PutObjectTaggingOutput, StreamingBlob, Tag, TagSet, Timestamp, TimestampFormat, + UploadPartInput, UploadPartOutput, }; use s3s::service::{S3Service, S3ServiceBuilder}; use s3s::validation::{AwsNameValidation, NameValidation}; @@ -88,6 +90,13 @@ const SOURCE_LEGALHOLD_TIMESTAMP_HEADERS: [&str; 2] = [ "x-rustfs-source-replication-legalhold-timestamp", "x-minio-source-replication-legalhold-timestamp", ]; +/// Wire prefix of the SSE-C passthrough replication transport headers +/// (`X-Rustfs-Replication-*`). In the default mode the fake stores them like a +/// RustFS target and echoes SSE-C evidence back on HEAD/GET; with +/// [`FakeS3Target::drop_unlisted_replication_headers`] it models MinIO / +/// generic S3, which silently discard unknown x-* headers. +const REPLICATION_SSE_TRANSPORT_PREFIX: &str = "x-rustfs-replication-"; +const REPLICATION_SSEC_ALGORITHM_TRANSPORT_HEADER: &str = "x-rustfs-replication-ssec-algorithm"; const RESERVED_BUCKET_PREFIXES: [&str; 3] = ["xn--", "sthree-", "amzn-s3-demo-"]; const RESERVED_BUCKET_SUFFIXES: [&str; 6] = ["-s3alias", "--ol-s3", ".mrap", "--x-s3", "--table-s3", "-an"]; @@ -103,6 +112,9 @@ pub enum Operation { GetObject, HeadObject, DeleteObject, + GetObjectTagging, + PutObjectTagging, + DeleteObjectTagging, ListObjectVersions, CreateMultipartUpload, UploadPart, @@ -149,6 +161,42 @@ impl ReplicationTimestampHeaders { } } +/// Read-proxy related headers observed on a request, journaled so proxy +/// tests can assert the exact wire contract: the anti-loop marker present, +/// the replication-check exemption absent, and the client SSE-C key family +/// forwarded verbatim. The SSE-C key value itself is never retained — only +/// its presence. +#[derive(Debug, Clone, Default, PartialEq, Eq)] +pub struct ProxyHeaderSnapshot { + pub source_proxy_request: Option, + pub replication_check: Option, + pub ssec_algorithm: Option, + pub ssec_key_present: bool, + pub ssec_key_md5: Option, + /// Whether the request carried any `X-Rustfs-Replication-*` SSE-C + /// passthrough transport header, so fail-closed tests can assert the + /// sender really shipped the material a dropping target discarded. + pub ssec_transport_present: bool, +} + +impl ProxyHeaderSnapshot { + fn from_headers(headers: &HeaderMap) -> Self { + Self { + source_proxy_request: header_value(headers, &["x-rustfs-source-proxy-request", "x-minio-source-proxy-request"]) + .map(bounded_journal_value), + replication_check: header_value(headers, &["x-rustfs-source-replication-check", "x-minio-source-replication-check"]) + .map(bounded_journal_value), + ssec_algorithm: header_value(headers, &["x-amz-server-side-encryption-customer-algorithm"]) + .map(bounded_journal_value), + ssec_key_present: headers.contains_key("x-amz-server-side-encryption-customer-key"), + ssec_key_md5: header_value(headers, &["x-amz-server-side-encryption-customer-key-md5"]).map(bounded_journal_value), + ssec_transport_present: headers + .keys() + .any(|name| name.as_str().starts_with(REPLICATION_SSE_TRANSPORT_PREFIX)), + } + } +} + /// Credential-free request metadata retained for deterministic assertions. #[derive(Debug, Clone, PartialEq, Eq)] pub struct RequestRecord { @@ -163,6 +211,7 @@ pub struct RequestRecord { pub content_length: Option, pub consumed_bytes: Option, pub replication_timestamps: ReplicationTimestampHeaders, + pub proxy_headers: ProxyHeaderSnapshot, pub fault: Option, } @@ -178,6 +227,10 @@ struct ControlState { struct StoreState { assign_own_version_ids: bool, assign_own_multipart_version_ids: bool, + /// MinIO-like mode: silently discard non-whitelisted replication + /// transport headers instead of storing them (see + /// [`REPLICATION_SSE_TRANSPORT_PREFIX`]). + drop_unlisted_replication_headers: bool, buckets: HashMap, uploads: HashMap, total_bytes: usize, @@ -199,6 +252,12 @@ struct ObjectVersion { delete_marker: bool, content_type: Option, metadata: Option>, + /// Object tags as ordered key/value pairs (PutObjectTagging replaces the + /// whole set, DeleteObjectTagging clears it). + tags: Vec<(String, String)>, + /// SSE-C passthrough transport headers stored with the version (RustFS + /// target behavior); empty when the drop mode discarded them. + replication_sse_headers: Vec<(String, String)>, } #[derive(Clone)] @@ -208,6 +267,7 @@ struct MultipartState { version_id: String, content_type: Option, metadata: Option>, + replication_sse_headers: Vec<(String, String)>, parts: BTreeMap, } @@ -428,6 +488,15 @@ impl FakeS3Target { /// Mint own version ids for the multipart path only — models a target /// that adopts PutObject version ids but not CreateMultipartUpload ones. + /// MinIO-like mode: silently drop every `X-Rustfs-Replication-*` SSE-C + /// passthrough transport header instead of storing it. The default (off) + /// models a RustFS target, which preserves the headers and echoes SSE-C + /// evidence (`x-amz-server-side-encryption-customer-algorithm`) on + /// HEAD/GET of the replica. + pub fn drop_unlisted_replication_headers(&self, enabled: bool) { + lock(&self.backend.store).drop_unlisted_replication_headers = enabled; + } + pub fn assign_own_multipart_version_ids(&self, enabled: bool) { lock(&self.backend.store).assign_own_multipart_version_ids = enabled; } @@ -569,6 +638,7 @@ impl S3Access for FaultAccess { .and_then(|value| value.to_str().ok()) .and_then(|value| value.parse().ok()); let replication_timestamps = ReplicationTimestampHeaders::from_headers(context.headers()); + let proxy_headers = ProxyHeaderSnapshot::from_headers(context.headers()); let fault = record_request( &self.control, operation, @@ -576,6 +646,7 @@ impl S3Access for FaultAccess { parsed, content_length, replication_timestamps, + proxy_headers, ); if let Some(RequestFault { action: FaultAction::Status(status), @@ -615,6 +686,9 @@ fn operation_from_s3_name(name: &str) -> Operation { "GetObject" => Operation::GetObject, "HeadObject" => Operation::HeadObject, "DeleteObject" => Operation::DeleteObject, + "GetObjectTagging" => Operation::GetObjectTagging, + "PutObjectTagging" => Operation::PutObjectTagging, + "DeleteObjectTagging" => Operation::DeleteObjectTagging, "CreateMultipartUpload" => Operation::CreateMultipartUpload, "UploadPart" => Operation::UploadPart, "CompleteMultipartUpload" => Operation::CompleteMultipartUpload, @@ -630,6 +704,7 @@ fn record_request( parsed: ParsedRequest, content_length: Option, replication_timestamps: ReplicationTimestampHeaders, + proxy_headers: ProxyHeaderSnapshot, ) -> Option { let mut state = lock(control); let action = parsed @@ -655,6 +730,7 @@ fn record_request( content_length, consumed_bytes: None, replication_timestamps, + proxy_headers, fault: action.clone(), }); action.map(|action| RequestFault { sequence, action }) @@ -721,6 +797,15 @@ fn parse_request(method: &Method, uri: &Uri) -> ParsedRequest { (&Method::POST, true) if query.contains_key("uploads") => Operation::CreateMultipartUpload, (&Method::POST, true) if upload_id.is_some() => Operation::CompleteMultipartUpload, (&Method::DELETE, true) if upload_id.is_some() => Operation::AbortMultipartUpload, + (&Method::GET, true) if query.contains_key("tagging") && only_query_keys(&["tagging", "versionId"]) => { + Operation::GetObjectTagging + } + (&Method::PUT, true) if query.contains_key("tagging") && only_query_keys(&["tagging", "versionId"]) => { + Operation::PutObjectTagging + } + (&Method::DELETE, true) if query.contains_key("tagging") && only_query_keys(&["tagging", "versionId"]) => { + Operation::DeleteObjectTagging + } // A replication PUT addresses the source version via `?versionId=`. (&Method::PUT, true) if only_query_keys(&["versionId"]) => Operation::PutObject, (&Method::GET, true) if only_query_keys(&["versionId"]) => Operation::GetObject, @@ -788,6 +873,29 @@ fn new_version_id(headers: &HeaderMap, assign_own: bool) -> S3Result { Ok(version_id.to_string()) } +/// Capture the SSE-C passthrough transport headers a replication PUT carried. +/// Returns an empty set in the MinIO-like drop mode. +fn captured_replication_sse_headers(headers: &HeaderMap, drop_unlisted: bool) -> Vec<(String, String)> { + if drop_unlisted { + return Vec::new(); + } + headers + .iter() + .filter(|(name, _)| name.as_str().starts_with(REPLICATION_SSE_TRANSPORT_PREFIX)) + .filter_map(|(name, value)| Some((name.as_str().to_string(), value.to_str().ok()?.to_string()))) + .collect() +} + +/// SSE-C evidence a RustFS-like target echoes for a stored passthrough +/// replica: the customer algorithm restored from the transport headers. +fn stored_sse_customer_algorithm(version: &ObjectVersion) -> Option { + version + .replication_sse_headers + .iter() + .find(|(name, _)| name == REPLICATION_SSEC_ALGORITHM_TRANSPORT_HEADER) + .map(|(_, value)| value.clone()) +} + fn source_etag(headers: &HeaderMap) -> S3Result> { header_value(headers, &SOURCE_ETAG_HEADERS) .map(|value| validate_retained_identifier(value, "source ETag").map(|value| normalize_etag(&value))) @@ -1135,6 +1243,33 @@ fn find_version(state: &StoreState, bucket: &str, key: &str, version_id: Option< Ok(version.clone()) } +/// Replace (or clear, with an empty vec) the tag set of the addressed +/// version, returning its version id. Mirrors `find_version` addressing: +/// explicit version id or the latest version, delete markers rejected. +fn set_version_tags( + state: &mut StoreState, + bucket: &str, + key: &str, + version_id: Option<&str>, + tags: Vec<(String, String)>, +) -> S3Result { + // Resolve first (immutable) so the error paths match find_version. + let resolved = find_version(state, bucket, key, version_id)?.version_id; + let versions = state + .buckets + .get_mut(bucket) + .expect("bucket existence checked by find_version") + .objects + .get_mut(key) + .expect("key existence checked by find_version"); + let version = versions + .iter_mut() + .find(|version| version.version_id == resolved) + .expect("version existence checked by find_version"); + version.tags = tags; + Ok(resolved) +} + #[async_trait] impl S3 for FakeBackend { async fn head_bucket(&self, req: S3Request) -> S3Result> { @@ -1231,7 +1366,10 @@ impl S3 for FakeBackend { let input = req.input; let body = collect_stream(input.body, input.content_length, fault.as_ref(), &self.control).await?; validate_stored_metadata(&input.content_type, &input.metadata)?; - let assign_own = lock(&self.store).assign_own_version_ids; + let (assign_own, drop_unlisted) = { + let state = lock(&self.store); + (state.assign_own_version_ids, state.drop_unlisted_replication_headers) + }; let version_id = new_version_id(&headers, assign_own)?; let e_tag = match source_etag(&headers)? { Some(value) => value, @@ -1248,6 +1386,8 @@ impl S3 for FakeBackend { delete_marker: false, content_type: input.content_type, metadata: input.metadata, + tags: Vec::new(), + replication_sse_headers: captured_replication_sse_headers(&headers, drop_unlisted), }; upsert_version(&mut lock(&self.store), &input.bucket, input.key, version)?; Ok(apply_response_fault( @@ -1268,6 +1408,7 @@ impl S3 for FakeBackend { let state = lock(&self.store); find_version(&state, &input.bucket, &input.key, input.version_id.as_deref())? }; + let sse_customer_algorithm = stored_sse_customer_algorithm(&version); Ok(apply_response_fault( S3Response::new(GetObjectOutput { body: Some(StreamingBlob::new(Body::from(version.body.clone()))), @@ -1277,6 +1418,7 @@ impl S3 for FakeBackend { e_tag: Some(ETag::Strong(version.e_tag)), last_modified: Some(version.last_modified.clone()), version_id: Some(version.version_id), + sse_customer_algorithm, ..Default::default() }), fault.as_ref(), @@ -1291,6 +1433,7 @@ impl S3 for FakeBackend { let state = lock(&self.store); find_version(&state, &input.bucket, &input.key, input.version_id.as_deref())? }; + let sse_customer_algorithm = stored_sse_customer_algorithm(&version); Ok(apply_response_fault( S3Response::new(HeadObjectOutput { content_length: Some(version.body.len() as i64), @@ -1299,12 +1442,79 @@ impl S3 for FakeBackend { e_tag: Some(ETag::Strong(version.e_tag)), last_modified: Some(version.last_modified.clone()), version_id: Some(version.version_id), + sse_customer_algorithm, ..Default::default() }), fault.as_ref(), )) } + async fn get_object_tagging(&self, req: S3Request) -> S3Result> { + let fault = request_fault(&req); + apply_non_body_fault(fault.as_ref(), &self.control).await?; + let input = req.input; + let version = { + let state = lock(&self.store); + find_version(&state, &input.bucket, &input.key, input.version_id.as_deref())? + }; + let tag_set: TagSet = version + .tags + .into_iter() + .map(|(key, value)| Tag { + key: Some(key), + value: Some(value), + }) + .collect(); + Ok(apply_response_fault( + S3Response::new(GetObjectTaggingOutput { + tag_set, + version_id: Some(ObjectVersionId::from(version.version_id)), + }), + fault.as_ref(), + )) + } + + async fn put_object_tagging(&self, req: S3Request) -> S3Result> { + let fault = request_fault(&req); + apply_non_body_fault(fault.as_ref(), &self.control).await?; + let input = req.input; + let tags = input + .tagging + .tag_set + .into_iter() + .map(|tag| (tag.key.unwrap_or_default(), tag.value.unwrap_or_default())) + .collect(); + let version_id = { + let mut state = lock(&self.store); + set_version_tags(&mut state, &input.bucket, &input.key, input.version_id.as_deref(), tags)? + }; + Ok(apply_response_fault( + S3Response::new(PutObjectTaggingOutput { + version_id: Some(ObjectVersionId::from(version_id)), + }), + fault.as_ref(), + )) + } + + async fn delete_object_tagging( + &self, + req: S3Request, + ) -> S3Result> { + let fault = request_fault(&req); + apply_non_body_fault(fault.as_ref(), &self.control).await?; + let input = req.input; + let version_id = { + let mut state = lock(&self.store); + set_version_tags(&mut state, &input.bucket, &input.key, input.version_id.as_deref(), Vec::new())? + }; + Ok(apply_response_fault( + S3Response::new(DeleteObjectTaggingOutput { + version_id: Some(ObjectVersionId::from(version_id)), + }), + fault.as_ref(), + )) + } + async fn delete_object(&self, req: S3Request) -> S3Result> { let fault = request_fault(&req); apply_non_body_fault(fault.as_ref(), &self.control).await?; @@ -1381,6 +1591,8 @@ impl S3 for FakeBackend { delete_marker: true, content_type: None, metadata: None, + tags: Vec::new(), + replication_sse_headers: Vec::new(), }, )?; Ok(apply_response_fault( @@ -1408,9 +1620,10 @@ impl S3 for FakeBackend { ensure_upload_budget(&state)?; validate_stored_metadata(&input.content_type, &input.metadata)?; let upload_id = Uuid::new_v4().to_string(); - // Read the flag before the mutable borrow of `state.uploads` below + // Read the flags before the mutable borrow of `state.uploads` below // (and never re-lock the store: the mutex is not reentrant). let mint_own = state.assign_own_version_ids || state.assign_own_multipart_version_ids; + let drop_unlisted = state.drop_unlisted_replication_headers; let version_id = new_version_id(&headers, mint_own)?; state.uploads.insert( upload_id.clone(), @@ -1420,6 +1633,7 @@ impl S3 for FakeBackend { version_id, content_type: input.content_type, metadata: input.metadata, + replication_sse_headers: captured_replication_sse_headers(&headers, drop_unlisted), parts: BTreeMap::new(), }, ); @@ -1557,6 +1771,7 @@ impl S3 for FakeBackend { version_id: upload.version_id.clone(), content_type: upload.content_type.clone(), metadata: upload.metadata.clone(), + replication_sse_headers: upload.replication_sse_headers.clone(), parts: BTreeMap::new(), }, selected, @@ -1583,6 +1798,8 @@ impl S3 for FakeBackend { delete_marker: false, content_type: upload.content_type, metadata: upload.metadata, + tags: Vec::new(), + replication_sse_headers: upload.replication_sse_headers, }; let mut state = lock(&self.store); let current = state @@ -1787,6 +2004,65 @@ mod tests { Ok(()) } + /// Default mode is RustFS-like: SSE-C passthrough transport headers are + /// stored and the customer algorithm is echoed on HEAD/GET. Drop mode is + /// MinIO-like: the headers are silently discarded, so no evidence comes + /// back — the exact difference the N2 fail-closed audit keys on. Both + /// modes journal that the sender shipped the transport headers. + #[tokio::test] + async fn ssec_passthrough_headers_echo_and_drop_modes() -> Result<(), BoxError> { + let target = FakeS3Target::start().await?; + target.create_bucket("target-bucket"); + let client = client(&target); + + let put_with_transport_headers = |key: &'static str| { + client + .put_object() + .bucket("target-bucket") + .key(key) + .body(ByteStream::from_static(b"ciphertext")) + .customize() + .map_request(move |mut request| { + let headers = request.headers_mut(); + headers.insert("x-rustfs-replication-ssec-algorithm", "AES256"); + headers.insert("x-rustfs-replication-ssec-key-md5", "AAAAAAAAAAAAAAAAAAAAAA=="); + Ok::<_, std::convert::Infallible>(request) + }) + .send() + }; + + put_with_transport_headers("kept").await?; + let head = client.head_object().bucket("target-bucket").key("kept").send().await?; + assert_eq!(head.sse_customer_algorithm(), Some("AES256")); + let get = client.get_object().bucket("target-bucket").key("kept").send().await?; + assert_eq!(get.sse_customer_algorithm(), Some("AES256")); + + target.drop_unlisted_replication_headers(true); + put_with_transport_headers("dropped").await?; + let head = client.head_object().bucket("target-bucket").key("dropped").send().await?; + assert_eq!(head.sse_customer_algorithm(), None, "drop mode must discard SSE-C evidence"); + + let requests = target.requests(); + for key in ["kept", "dropped"] { + let record = requests + .iter() + .find(|record| record.operation == Operation::PutObject && record.key.as_deref() == Some(key)) + .expect("PUT must be journaled"); + assert!( + record.proxy_headers.ssec_transport_present, + "the journal must prove the sender shipped the transport headers for {key}" + ); + } + let plain_head = requests + .iter() + .find(|record| record.operation == Operation::HeadObject) + .expect("HEAD must be journaled"); + assert!(!plain_head.proxy_headers.ssec_transport_present); + + target.shutdown().await; + Ok(()) + } + macro_rules! assert_sdk_error { ($error:expr, $status:expr, $code:expr) => {{ let error = &$error; @@ -3052,6 +3328,7 @@ mod tests { version_id: index.to_string(), content_type: None, metadata: None, + replication_sse_headers: Vec::new(), parts: BTreeMap::new(), }, ); @@ -3074,6 +3351,7 @@ mod tests { }, Some(0), ReplicationTimestampHeaders::default(), + ProxyHeaderSnapshot::default(), ); } let records = lock(&control).requests.clone(); @@ -3096,6 +3374,7 @@ mod tests { }, None, ReplicationTimestampHeaders::default(), + ProxyHeaderSnapshot::default(), ); { let bounded_records = lock(&bounded_control); diff --git a/crates/e2e_test/src/replication_extension_test.rs b/crates/e2e_test/src/replication_extension_test.rs index ec847bfb7..8279479f1 100644 --- a/crates/e2e_test/src/replication_extension_test.rs +++ b/crates/e2e_test/src/replication_extension_test.rs @@ -2606,17 +2606,20 @@ async fn test_replication_check_succeeds_with_remote_target() -> Result<(), Box< assert_eq!(response.status(), StatusCode::OK); let payload: serde_json::Value = response.json().await?; - assert_eq!(payload["Status"], "OK"); + assert_eq!(payload["Status"], "OK", "{payload}"); assert_eq!(payload["ActiveMutation"], true); assert_eq!(payload["Targets"].as_array().map(Vec::len), Some(1)); - assert_eq!(payload["Targets"][0]["Status"], "OK"); - assert_eq!(payload["Targets"][0]["Phases"]["Put"]["Status"], "OK"); + assert_eq!(payload["Targets"][0]["Status"], "OK", "{payload}"); + assert_eq!(payload["Targets"][0]["Phases"]["Put"]["Status"], "OK", "{payload}"); // A RustFS target adopts the source version id, so the P1-19 // version-identity probe passes. - assert_eq!(payload["Targets"][0]["Phases"]["VersionFidelity"]["Status"], "OK"); - assert_eq!(payload["Targets"][0]["Phases"]["DeleteMarker"]["Status"], "OK"); - assert_eq!(payload["Targets"][0]["Phases"]["VersionDelete"]["Status"], "OK"); - assert_eq!(payload["Targets"][0]["Phases"]["Cleanup"]["Status"], "OK"); + assert_eq!(payload["Targets"][0]["Phases"]["VersionFidelity"]["Status"], "OK", "{payload}"); + // A RustFS target preserves the SSE-C passthrough transport headers and + // echoes the customer algorithm on the replication-check HEAD (N2). + assert_eq!(payload["Targets"][0]["Phases"]["SsecPassthrough"]["Status"], "OK", "{payload}"); + assert_eq!(payload["Targets"][0]["Phases"]["DeleteMarker"]["Status"], "OK", "{payload}"); + assert_eq!(payload["Targets"][0]["Phases"]["VersionDelete"]["Status"], "OK", "{payload}"); + assert_eq!(payload["Targets"][0]["Phases"]["Cleanup"]["Status"], "OK", "{payload}"); let target_client = target_env.create_s3_client(); let versions = target_client @@ -4616,6 +4619,410 @@ async fn test_bucket_replication_sse_c_multipart_passthrough() -> TestResult { Ok(()) } +/// N2 (backlog#1675 P1-22): SSE-C passthrough replication to a target that +/// silently drops the `X-Rustfs-Replication-*` transport headers (MinIO-like +/// behavior, modeled by the fake target's drop mode) used to report COMPLETED +/// while the replica had irrecoverably lost its decryption material — the red +/// light this test was born failing on. Fail-closed contract now under test: +/// the first attempt PUTs, HEAD-backs the replica, finds no SSE-C evidence, +/// records the target Unsupported and reports FAILED; a second SSE-C object +/// fails without any PUT reaching the target (capability cache, proven from +/// the target journal); plaintext objects still replicate COMPLETED. +#[tokio::test] +#[serial] +async fn test_ssec_replication_fails_closed_when_target_drops_passthrough_headers() -> TestResult { + init_logging(); + + let target = FakeS3Target::start().await?; + let target_bucket = "ssec-drop-dst"; + target.create_bucket(target_bucket); + target.drop_unlisted_replication_headers(true); + + let mut source_env = RustFSTestEnvironment::new().await?; + let mut env_vars = replication_fast_env(); + env_vars.extend_from_slice(LOOPBACK_REPLICATION_TARGET_ENV); + env_vars.extend_from_slice(&[("NO_PROXY", "127.0.0.1,localhost"), ("HTTP_PROXY", ""), ("HTTPS_PROXY", "")]); + source_env.start_rustfs_server_with_env(vec![], &env_vars).await?; + + let source_bucket = "ssec-drop-src"; + let source_client = source_env.create_s3_client(); + source_client.create_bucket().bucket(source_bucket).send().await?; + enable_bucket_versioning(&source_env, source_bucket).await?; + let target_arn = set_replication_target_with_options( + &source_env, + source_bucket, + ReplicationTargetOptions { + endpoint: &target.address(), + access_key: FAKE_ACCESS_KEY, + secret_key: FAKE_SECRET_KEY, + target_bucket, + secure: false, + skip_tls_verify: false, + ca_cert_pem: None, + }, + ) + .await?; + put_bucket_replication(&source_env, source_bucket, &target_arn).await?; + + let customer_key = BASE64_STANDARD.encode(REPL17_SSEC_KEY); + let customer_key_md5 = sse_customer_key_md5_base64(REPL17_SSEC_KEY); + let put_ssec = |key: &'static str| { + source_client + .put_object() + .bucket(source_bucket) + .key(key) + .body(ByteStream::from_static(b"ssec fail-closed payload")) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + }; + + // First SSE-C object: the audit must catch the dropped material. + put_ssec("ssec-first.txt").await?; + wait_for_source_replication_status(&source_client, source_bucket, "ssec-first.txt", "FAILED", true).await?; + + let requests = target.take_requests(); + let first_put = requests + .iter() + .find(|record| record.operation == FakeTargetOperation::PutObject && record.key.as_deref() == Some("ssec-first.txt")) + .ok_or("the first SSE-C object must have been PUT (capability was Unknown)")?; + assert!( + first_put.proxy_headers.ssec_transport_present, + "the replication PUT must have shipped the SSE-C transport headers the target then dropped" + ); + assert!( + requests.iter().any(|record| { + record.operation == FakeTargetOperation::HeadObject + && record.key.as_deref() == Some("ssec-first.txt") + && record.sequence > first_put.sequence + && record.proxy_headers.replication_check.as_deref() == Some("true") + }), + "the post-PUT HEAD-back audit must have run through the replication-check channel; journal: {requests:?}" + ); + + // Second SSE-C object: the cached Unsupported verdict fails it closed + // before any PUT — including MRF retries of the first object. + put_ssec("ssec-second.txt").await?; + wait_for_source_replication_status(&source_client, source_bucket, "ssec-second.txt", "FAILED", true).await?; + assert!( + !target.requests().iter().any(|record| { + record.operation == FakeTargetOperation::PutObject + && record.key.as_deref() != Some("plain-control.txt") + && record.proxy_headers.ssec_transport_present + }), + "no further SSE-C ciphertext may reach a target recorded Unsupported; journal: {:?}", + target.requests() + ); + + // The gate is scoped to SSE-C: plaintext replication keeps working. + source_client + .put_object() + .bucket(source_bucket) + .key("plain-control.txt") + .body(ByteStream::from_static(b"plaintext control payload")) + .send() + .await?; + wait_for_source_replication_status(&source_client, source_bucket, "plain-control.txt", "COMPLETED", false).await?; + assert!(target.has_object(target_bucket, "plain-control.txt")); + + target.shutdown().await; + Ok(()) +} + +/// N2 (backlog#1675 P1-22): the admin replication-check must expose the same +/// verdict operators would otherwise only learn from failing SSE-C objects — +/// an SsecPassthrough probe phase that fails with the machine-readable +/// `BucketRemoteSsecPassthroughUnsupported` code against a header-dropping +/// target, with no probe residue left behind. The target's overall status +/// stays OK: unlike version-identity drift, dropped passthrough headers are +/// a capability limit, and a plaintext-only deployment against a MinIO-like +/// target must not turn red. +#[tokio::test] +#[serial] +async fn test_replication_check_flags_ssec_passthrough_dropping_target() -> TestResult { + init_logging(); + + let target = FakeS3Target::start().await?; + let target_bucket = "ssec-check-dst"; + target.create_bucket(target_bucket); + target.drop_unlisted_replication_headers(true); + + let mut source_env = RustFSTestEnvironment::new().await?; + let mut env_vars = replication_fast_env(); + env_vars.extend_from_slice(LOOPBACK_REPLICATION_TARGET_ENV); + env_vars.extend_from_slice(&[("NO_PROXY", "127.0.0.1,localhost"), ("HTTP_PROXY", ""), ("HTTPS_PROXY", "")]); + source_env.start_rustfs_server_with_env(vec![], &env_vars).await?; + + let source_bucket = "ssec-check-src"; + let source_client = source_env.create_s3_client(); + source_client.create_bucket().bucket(source_bucket).send().await?; + enable_bucket_versioning(&source_env, source_bucket).await?; + let target_arn = set_replication_target_with_options( + &source_env, + source_bucket, + ReplicationTargetOptions { + endpoint: &target.address(), + access_key: FAKE_ACCESS_KEY, + secret_key: FAKE_SECRET_KEY, + target_bucket, + secure: false, + skip_tls_verify: false, + ca_cert_pem: None, + }, + ) + .await?; + put_bucket_replication(&source_env, source_bucket, &target_arn).await?; + + let response = run_replication_check(&source_env, source_bucket).await?; + assert_eq!(response.status(), StatusCode::OK); + let payload: serde_json::Value = response.json().await?; + + assert_eq!( + payload["Status"], "OK", + "a capability-only SSE-C failure must not fail the check overall: {payload}" + ); + let target_report = &payload["Targets"][0]; + assert_eq!(target_report["Status"], "OK", "{payload}"); + let ssec = &target_report["Phases"]["SsecPassthrough"]; + assert_eq!(ssec["Status"], "FAILED", "SsecPassthrough phase must fail: {payload}"); + assert_eq!( + ssec["Code"], "BucketRemoteSsecPassthroughUnsupported", + "the failure must carry the machine-readable code: {payload}" + ); + // Basic replication of plaintext objects works on this target: every other + // phase passes, so the code is the discriminator operators branch on. + assert_eq!(target_report["Phases"]["Put"]["Status"], "OK", "{payload}"); + assert_eq!(target_report["Phases"]["VersionFidelity"]["Status"], "OK", "{payload}"); + assert_eq!(target_report["Phases"]["DeleteMarker"]["Status"], "OK", "{payload}"); + assert_eq!(target_report["Phases"]["VersionDelete"]["Status"], "OK", "{payload}"); + assert_eq!(target_report["Phases"]["Cleanup"]["Status"], "OK", "{payload}"); + + // The SSE-C probe PUT must have shipped the real transport header names — + // a mangled or missing header set would fail the phase for the wrong + // reason and mask a working target. + let requests = target.requests(); + assert!( + requests + .iter() + .any(|record| record.operation == FakeTargetOperation::PutObject && record.proxy_headers.ssec_transport_present), + "the SSE-C probe PUT must carry the X-Rustfs-Replication-* transport headers; journal: {requests:?}" + ); + + // No probe residue, including the SSE-C probe version. + let probe_put = requests + .into_iter() + .find(|record| record.operation == FakeTargetOperation::PutObject) + .ok_or("the probe PUT never reached the fake target")?; + let probe_key = probe_put.key.ok_or("probe PUT journal record has no key")?; + assert!( + target.stored_versions(target_bucket, &probe_key).is_empty(), + "all probe versions must be cleaned up" + ); + + target.shutdown().await; + Ok(()) +} + +/// C1 (backlog#1675 P1-22): heal-path convergence for SSE-C. An SSE-C object +/// whose live replication failed during a target outage must converge through +/// the scanner/heal compensation once the target returns — passing the N2 +/// HEAD-back audit against the recovered RustFS target — and the replica must +/// be readable with the customer key. +#[tokio::test] +#[serial] +async fn test_bucket_replication_sse_c_heals_after_target_outage() -> TestResult { + init_logging(); + + let (source_env, mut target_env, source_bucket, target_bucket) = + build_sse_replication_pair("ssec-heal", false, false).await?; + let source_client = source_env.create_s3_client(); + let key = "ssec-heal-contract.txt"; + let body = b"repl-22 ssec heal payload".to_vec(); + let customer_key = BASE64_STANDARD.encode(REPL17_SSEC_KEY); + let customer_key_md5 = sse_customer_key_md5_base64(REPL17_SSEC_KEY); + + // Target outage: the SSE-C write cannot replicate. + target_env.stop_server(); + + source_client + .put_object() + .bucket(&source_bucket) + .key(key) + .body(ByteStream::from(body.clone())) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + + // The failure is observable on the source (SSE-C HEAD needs the key). + let deadline = tokio::time::Instant::now() + Duration::from_secs(30); + loop { + let head = source_client + .head_object() + .bucket(&source_bucket) + .key(key) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + match head.replication_status().map(|status| status.as_str()) { + Some("PENDING") | Some("FAILED") => break, + other => { + if tokio::time::Instant::now() >= deadline { + return Err(format!("source SSE-C object never reported PENDING/FAILED; last status={other:?}").into()); + } + sleep(Duration::from_millis(200)).await; + } + } + } + + // Recover the target in place; the source scanner re-drives the failure. + target_env + .restart_server_preserving_data(vec![], &[("NO_PROXY", "127.0.0.1,localhost"), ("HTTP_PROXY", ""), ("HTTPS_PROXY", "")]) + .await?; + + wait_for_source_replication_status(&source_client, &source_bucket, key, "COMPLETED", true).await?; + + // The healed replica is a REPLICA (status surfaces on HEAD) readable with + // the customer key. + let target_client = target_env.create_s3_client(); + let replica_head = target_client + .head_object() + .bucket(&target_bucket) + .key(key) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + assert_eq!( + replica_head.replication_status().map(|status| status.as_str()), + Some("REPLICA"), + "the healed copy must carry REPLICA status" + ); + let replica = target_client + .get_object() + .bucket(&target_bucket) + .key(key) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + assert_eq!(replica.sse_customer_algorithm(), Some("AES256")); + assert_eq!(replica.body.collect().await?.into_bytes().as_ref(), body.as_slice()); + + Ok(()) +} + +/// C1 (backlog#1675 P1-22): existing-object resync for SSE-C. An SSE-C object +/// written BEFORE any replication config must reach the RustFS target through +/// the existing-object resync (`replicate_all` transport, N2-audited), land as +/// a REPLICA, and read back with the customer key. +#[tokio::test] +#[serial] +async fn test_bucket_replication_sse_c_existing_object_resync() -> TestResult { + init_logging(); + + let mut source_env = RustFSTestEnvironment::new().await?; + let mut source_process_env = replication_fast_env(); + source_process_env.extend_from_slice(LOOPBACK_REPLICATION_TARGET_ENV); + source_process_env.extend_from_slice(FAST_SCANNER_ENV); + source_process_env.extend_from_slice(&[("NO_PROXY", "127.0.0.1,localhost"), ("HTTP_PROXY", ""), ("HTTPS_PROXY", "")]); + source_env.start_rustfs_server_with_env(vec![], &source_process_env).await?; + + let mut target_env = RustFSTestEnvironment::new().await?; + target_env + .start_rustfs_server_without_cleanup_with_env(&[ + ("NO_PROXY", "127.0.0.1,localhost"), + ("HTTP_PROXY", ""), + ("HTTPS_PROXY", ""), + ]) + .await?; + + let source_bucket = "ssec-existing-src"; + let target_bucket = "ssec-existing-dst"; + let source_client = source_env.create_s3_client(); + let target_client = target_env.create_s3_client(); + source_client.create_bucket().bucket(source_bucket).send().await?; + target_client.create_bucket().bucket(target_bucket).send().await?; + enable_bucket_versioning(&source_env, source_bucket).await?; + enable_bucket_versioning(&target_env, target_bucket).await?; + + // The SSE-C object exists before any replication wiring. + let key = "ssec-existing-contract.txt"; + let body = b"repl-22 ssec existing-object payload".to_vec(); + let customer_key = BASE64_STANDARD.encode(REPL17_SSEC_KEY); + let customer_key_md5 = sse_customer_key_md5_base64(REPL17_SSEC_KEY); + source_client + .put_object() + .bucket(source_bucket) + .key(key) + .body(ByteStream::from(body.clone())) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + + // Wire replication (existing-object enabled) and drive a resync. + let target_arn = set_replication_target(&source_env, source_bucket, &target_env, target_bucket).await?; + put_bucket_replication(&source_env, source_bucket, &target_arn).await?; + let (reset_arn, reset_id) = start_bucket_replication_reset(&source_env, source_bucket).await?; + assert_eq!(reset_arn, target_arn); + let terminal = wait_for_replication_reset_target(&source_env, source_bucket, &target_arn, |status| { + status.reset_id == reset_id && matches!(status.status.as_str(), "Completed" | "Failed") + }) + .await?; + assert_eq!(terminal.status, "Completed", "SSE-C existing-object resync must complete"); + assert!(terminal.replicated_count >= 1, "the existing SSE-C object must have been resynced"); + + // The replica is a REPLICA (status surfaces on HEAD) readable with the + // customer key. + let replica_head = target_client + .head_object() + .bucket(target_bucket) + .key(key) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + assert_eq!( + replica_head.replication_status().map(|status| status.as_str()), + Some("REPLICA"), + "the resynced copy must carry REPLICA status" + ); + let replica = target_client + .get_object() + .bucket(target_bucket) + .key(key) + .sse_customer_algorithm("AES256") + .sse_customer_key(&customer_key) + .sse_customer_key_md5(&customer_key_md5) + .send() + .await?; + assert_eq!(replica.sse_customer_algorithm(), Some("AES256")); + assert_eq!(replica.body.collect().await?.into_bytes().as_ref(), body.as_slice()); + + // No plaintext leak: the replica stays unreadable without the key. + assert!( + target_client + .get_object() + .bucket(target_bucket) + .key(key) + .send() + .await + .is_err(), + "SSE-C replica must not be readable without the customer key" + ); + + Ok(()) +} + /// backlog#1147 repl-17 / backlog#1783: SSE-S3 objects replicate by decrypting /// at the source and re-encrypting on the target with the target's own KMS. /// The property backlog#1291 pinned — never a silent plaintext replica — still @@ -8348,3 +8755,304 @@ async fn test_scanner_never_compensates_when_existing_object_replication_disable Ok(()) } + +/// Shared setup for the P1-5 read-proxy scenarios (backlog#1675): a RustFS +/// source with an enabled replication rule pointing at the fake target, and +/// an object seeded DIRECTLY on the target — it exists remotely but not +/// locally, exactly the active-active replication-lag window the read proxy +/// serves. +async fn start_read_proxy_lab( + source_bucket: &str, + target_bucket: &str, +) -> Result<(FakeS3Target, RustFSTestEnvironment, Client, Client), Box> { + let target = FakeS3Target::start().await?; + target.create_bucket(target_bucket); + target.assign_own_version_ids(true); + + let mut source_env = RustFSTestEnvironment::new().await?; + let mut process_env = replication_fast_env(); + process_env.extend_from_slice(LOOPBACK_REPLICATION_TARGET_ENV); + process_env.extend_from_slice(&[ + ("NO_PROXY", "127.0.0.1,localhost"), + ("HTTP_PROXY", ""), + ("HTTPS_PROXY", ""), + ("RUST_LOG", "error"), + ]); + source_env.start_rustfs_server_with_env(vec![], &process_env).await?; + + let source_client = source_env.create_s3_client(); + source_client.create_bucket().bucket(source_bucket).send().await?; + enable_bucket_versioning(&source_env, source_bucket).await?; + let target_arn = set_replication_target_with_options( + &source_env, + source_bucket, + ReplicationTargetOptions { + endpoint: &target.address(), + access_key: FAKE_ACCESS_KEY, + secret_key: FAKE_SECRET_KEY, + target_bucket, + secure: false, + skip_tls_verify: false, + ca_cert_pem: None, + }, + ) + .await?; + put_bucket_replication(&source_env, source_bucket, &target_arn).await?; + + let target_client = Client::from_conf(crate::common::build_test_s3_config( + target.endpoint(), + FAKE_ACCESS_KEY, + FAKE_SECRET_KEY, + None, + "read-proxy-e2e", + )); + + Ok((target, source_env, source_client, target_client)) +} + +/// P1-5 (backlog#1675): during the active-active replication lag window a +/// GET/HEAD for an object the local site does not have yet is proxied to the +/// replication target. Pins the wire contract: the anti-loop +/// `source-proxy-request` marker is sent, the replication worker's +/// `source-replication-check` SSE-C exemption is NEVER sent, client SSE-C +/// headers are forwarded verbatim, and an inbound request that was itself +/// proxied is answered locally (404) without touching the target. +#[tokio::test] +#[serial] +async fn test_get_and_head_proxy_unreplicated_object_to_replication_target() -> TestResult { + init_logging(); + + let source_bucket = "proxy-read-src"; + let target_bucket = "proxy-read-dst"; + let (target, source_env, source_client, target_client) = start_read_proxy_lab(source_bucket, target_bucket).await?; + + let payload = b"proxy payload".to_vec(); + target_client + .put_object() + .bucket(target_bucket) + .key("proxy-only") + .body(ByteStream::from(payload.clone())) + .send() + .await?; + target.take_requests(); + + // a. GET of the locally-missing object is served through the proxy. + let got = source_client + .get_object() + .bucket(source_bucket) + .key("proxy-only") + .send() + .await + .map_err(|err| format!("proxied GET failed: {}", err.into_service_error()))?; + assert_eq!(got.content_length, Some(payload.len() as i64)); + let body = got.body.collect().await?.into_bytes(); + assert_eq!(body.as_ref(), payload.as_slice(), "proxied GET must stream the target's body"); + + let get_record = target + .requests() + .into_iter() + .find(|record| record.operation == FakeTargetOperation::GetObject && record.key.as_deref() == Some("proxy-only")) + .ok_or("fake target never received the proxied GET")?; + assert_eq!( + get_record.proxy_headers.source_proxy_request.as_deref(), + Some("true"), + "proxied GET must carry the anti-loop source-proxy-request marker" + ); + assert!( + get_record.proxy_headers.replication_check.is_none(), + "proxied GET must never carry the replication worker's source-replication-check exemption" + ); + assert!( + get_record.proxy_headers.ssec_algorithm.is_none() && !get_record.proxy_headers.ssec_key_present, + "no client SSE-C headers were sent, so none may be forwarded" + ); + + // a2. Client SSE-C headers travel verbatim to the target (the target owns + // the real SSE-C decryption; the plaintext fake simply ignores them). + target.take_requests(); + let ssec_key = "01234567890123456789012345678901"; + let ssec_key_b64 = BASE64_STANDARD.encode(ssec_key); + let ssec_key_md5 = sse_customer_key_md5_base64(ssec_key); + let _ = source_client + .get_object() + .bucket(source_bucket) + .key("proxy-only") + .sse_customer_algorithm("AES256") + .sse_customer_key(&ssec_key_b64) + .sse_customer_key_md5(&ssec_key_md5) + .send() + .await + .map_err(|err| format!("proxied SSE-C GET failed: {}", err.into_service_error()))?; + let ssec_record = target + .requests() + .into_iter() + .find(|record| record.operation == FakeTargetOperation::GetObject && record.key.as_deref() == Some("proxy-only")) + .ok_or("fake target never received the proxied SSE-C GET")?; + assert_eq!(ssec_record.proxy_headers.ssec_algorithm.as_deref(), Some("AES256")); + assert!(ssec_record.proxy_headers.ssec_key_present, "SSE-C key header must be forwarded verbatim"); + assert_eq!(ssec_record.proxy_headers.ssec_key_md5.as_deref(), Some(ssec_key_md5.as_str())); + assert!(ssec_record.proxy_headers.replication_check.is_none()); + + // b. HEAD of the locally-missing object is served through the proxy. + target.take_requests(); + let head = source_client + .head_object() + .bucket(source_bucket) + .key("proxy-only") + .send() + .await + .map_err(|err| format!("proxied HEAD failed: {}", err.into_service_error()))?; + assert_eq!(head.content_length, Some(payload.len() as i64)); + let head_record = target + .requests() + .into_iter() + .find(|record| record.operation == FakeTargetOperation::HeadObject && record.key.as_deref() == Some("proxy-only")) + .ok_or("fake target never received the proxied HEAD")?; + assert_eq!(head_record.proxy_headers.source_proxy_request.as_deref(), Some("true")); + assert!(head_record.proxy_headers.replication_check.is_none()); + + // c. Anti-loop: an inbound request that already carries the proxy marker + // is answered locally with 404 and never forwarded to the target. + target.take_requests(); + let err = source_client + .get_object() + .bucket(source_bucket) + .key("proxy-only") + .customize() + .mutate_request(|req| { + req.headers_mut().insert("x-minio-source-proxy-request", "true"); + }) + .send() + .await + .expect_err("anti-loop GET must fail locally instead of proxying"); + let service_err = err.into_service_error(); + assert!(service_err.is_no_such_key(), "anti-loop GET must 404, got: {service_err}"); + assert!( + !target + .requests() + .iter() + .any(|record| record.operation == FakeTargetOperation::GetObject), + "anti-loop GET must not reach the replication target; journal: {:?}", + target.requests() + ); + + // c2. MinIO ProxyHeaderSet parity: the header's mere PRESENCE disables + // proxying — "false" is exactly what a peer's replication worker sends on + // its convergence HEADs, and proxying that miss back would fake + // convergence. + target.take_requests(); + let err = source_client + .get_object() + .bucket(source_bucket) + .key("proxy-only") + .customize() + .mutate_request(|req| { + req.headers_mut().insert("x-minio-source-proxy-request", "false"); + }) + .send() + .await + .expect_err("proxy-header-set GET must fail locally instead of proxying"); + let service_err = err.into_service_error(); + assert!(service_err.is_no_such_key(), "proxy-header-set GET must 404, got: {service_err}"); + assert!( + !target + .requests() + .iter() + .any(|record| record.operation == FakeTargetOperation::GetObject), + "proxy-header-set GET must not reach the replication target; journal: {:?}", + target.requests() + ); + + // d. The replication worker's own convergence HEAD against the target + // must carry `source-proxy-request: false` (never proxied back) and the + // replication-check exemption. Trigger real replication and inspect the + // fake journal. + target.take_requests(); + source_client + .put_object() + .bucket(source_bucket) + .key("worker-replicated") + .body(ByteStream::from_static(b"worker payload")) + .send() + .await?; + wait_for_target_request_version_id(&target, FakeTargetOperation::PutObject, "worker-replicated").await?; + let worker_head = target + .requests() + .into_iter() + .find(|record| record.operation == FakeTargetOperation::HeadObject && record.key.as_deref() == Some("worker-replicated")) + .ok_or_else(|| format!("replication worker never HEAD-ed the target; journal: {:?}", target.requests()))?; + assert_eq!( + worker_head.proxy_headers.source_proxy_request.as_deref(), + Some("false"), + "worker convergence HEAD must send source-proxy-request: false so the target answers locally" + ); + assert_eq!( + worker_head.proxy_headers.replication_check.as_deref(), + Some("true"), + "worker convergence HEAD keeps the replication-check exemption" + ); + + drop(source_env); + target.shutdown().await; + Ok(()) +} + +/// P1-5 (backlog#1675): GetObjectTagging for an object missing locally is +/// proxied to the replication target with the anti-loop marker, mirroring +/// MinIO `proxyGetTaggingToRepTarget`. +#[tokio::test] +#[serial] +async fn test_get_object_tagging_proxies_unreplicated_object_to_replication_target() -> TestResult { + init_logging(); + + let source_bucket = "proxy-tag-src"; + let target_bucket = "proxy-tag-dst"; + let (target, source_env, source_client, target_client) = start_read_proxy_lab(source_bucket, target_bucket).await?; + + target_client + .put_object() + .bucket(target_bucket) + .key("proxy-tagged") + .body(ByteStream::from_static(b"tagged payload")) + .send() + .await?; + target_client + .put_object_tagging() + .bucket(target_bucket) + .key("proxy-tagged") + .tagging( + aws_sdk_s3::types::Tagging::builder() + .tag_set(aws_sdk_s3::types::Tag::builder().key("team").value("storage").build()?) + .build()?, + ) + .send() + .await?; + target.take_requests(); + + let tags = source_client + .get_object_tagging() + .bucket(source_bucket) + .key("proxy-tagged") + .send() + .await + .map_err(|err| format!("proxied GetObjectTagging failed: {}", err.into_service_error()))?; + assert_eq!(tags.tag_set.len(), 1, "proxied tagging read must return the target's tags"); + assert_eq!(tags.tag_set[0].key.as_str(), "team"); + assert_eq!(tags.tag_set[0].value.as_str(), "storage"); + + let record = target + .requests() + .into_iter() + .find(|record| record.operation == FakeTargetOperation::GetObjectTagging && record.key.as_deref() == Some("proxy-tagged")) + .ok_or("fake target never received the proxied GetObjectTagging")?; + assert_eq!( + record.proxy_headers.source_proxy_request.as_deref(), + Some("true"), + "proxied tagging read must carry the anti-loop marker" + ); + assert!(record.proxy_headers.replication_check.is_none()); + + drop(source_env); + target.shutdown().await; + Ok(()) +} diff --git a/crates/ecstore/src/api/mod.rs b/crates/ecstore/src/api/mod.rs index 3a031d60a..17fffac3d 100644 --- a/crates/ecstore/src/api/mod.rs +++ b/crates/ecstore/src/api/mod.rs @@ -32,7 +32,7 @@ pub mod bucket { pub mod bucket_target_sys { pub use crate::bucket::bucket_target_sys::{ AdvancedPutOptions, BucketTargetError, BucketTargetSys, PutObjectOptions, RemoveObjectOptions, S3ClientError, - TargetClient, append_version_id_query, + SsecPassthroughCapability, TargetClient, append_version_id_query, }; } @@ -198,12 +198,13 @@ pub mod bucket { ReplicationType, ResyncOpts, ResyncStatusType, RuntimeReplicationTargetBacklog, TargetReplicationResyncStatus, VersionPurgeStatusType, XferStats, commit_force_delete_intent, complete_force_delete_intent, delete_replication_state_from_config, delete_replication_version_id, get_global_replication_pool, - get_global_replication_stats, init_background_replication, invalid_replication_config_status_field, - persist_force_delete_intent, read_durable_mrf_backlog, replication_state_to_filemeta, replication_status_to_filemeta, - replication_statuses_map, replication_target_arns, resync_start_conflict_id, should_remove_replication_target, - should_schedule_delete_replication, should_use_existing_delete_replication_info, - should_use_existing_delete_replication_source, unsupported_replication_config_field, - validate_replication_config_structure, validate_replication_config_target_arns, version_purge_status_to_filemeta, + get_global_replication_stats, get_proxy_targets, init_background_replication, + invalid_replication_config_status_field, persist_force_delete_intent, read_durable_mrf_backlog, + replication_state_to_filemeta, replication_status_to_filemeta, replication_statuses_map, replication_target_arns, + resync_start_conflict_id, should_remove_replication_target, should_schedule_delete_replication, + should_use_existing_delete_replication_info, should_use_existing_delete_replication_source, + unsupported_replication_config_field, validate_replication_config_structure, validate_replication_config_target_arns, + version_purge_status_to_filemeta, }; } diff --git a/crates/ecstore/src/bucket/bucket_target_sys.rs b/crates/ecstore/src/bucket/bucket_target_sys.rs index 00cf124f1..ef1918648 100644 --- a/crates/ecstore/src/bucket/bucket_target_sys.rs +++ b/crates/ecstore/src/bucket/bucket_target_sys.rs @@ -27,10 +27,15 @@ use aws_sdk_s3::config::SharedHttpClient; use aws_sdk_s3::error::ProvideErrorMetadata; use aws_sdk_s3::error::SdkError; use aws_sdk_s3::operation::complete_multipart_upload::CompleteMultipartUploadOutput; +use aws_sdk_s3::operation::delete_object_tagging::{DeleteObjectTaggingError, DeleteObjectTaggingOutput}; +use aws_sdk_s3::operation::get_object::{GetObjectError, GetObjectOutput}; +use aws_sdk_s3::operation::get_object_tagging::{GetObjectTaggingError, GetObjectTaggingOutput}; use aws_sdk_s3::operation::head_bucket::HeadBucketError; use aws_sdk_s3::operation::head_object::HeadObjectError; +use aws_sdk_s3::operation::put_object_tagging::{PutObjectTaggingError, PutObjectTaggingOutput}; use aws_sdk_s3::operation::upload_part::UploadPartOutput; use aws_sdk_s3::primitives::ByteStream; +use aws_sdk_s3::types::Tagging as SdkTagging; use aws_sdk_s3::types::{ ChecksumMode, CompletedMultipartUpload, CompletedPart, ObjectLockLegalHoldStatus, ObjectLockRetentionMode, }; @@ -57,8 +62,8 @@ use rustfs_utils::http::{ is_rustfs_header, is_standard_header, is_storageclass_header, }; use rustfs_utils::http::{ - SUFFIX_FORCE_DELETE, SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_ETAG, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_REPLICATION_CHECK, - SUFFIX_SOURCE_REPLICATION_LEGALHOLD_TIMESTAMP, SUFFIX_SOURCE_REPLICATION_REQUEST, + SUFFIX_FORCE_DELETE, SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_ETAG, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_PROXY_REQUEST, + SUFFIX_SOURCE_REPLICATION_CHECK, SUFFIX_SOURCE_REPLICATION_LEGALHOLD_TIMESTAMP, SUFFIX_SOURCE_REPLICATION_REQUEST, SUFFIX_SOURCE_REPLICATION_RETENTION_TIMESTAMP, SUFFIX_SOURCE_REPLICATION_TAGGING_TIMESTAMP, SUFFIX_SOURCE_VERSION_ID, insert_header, }; @@ -294,9 +299,41 @@ struct TargetClientBuildProbe { release: Arc, } +/// SSE-C passthrough capability verdicts (see the enum's own docs in +/// `rustfs-replication`) are cached here per target ARN: entries follow the +/// `arn_remotes_map` lifecycle (rebuilding or removing a target resets its +/// capability to `Unknown`) and additionally expire after +/// [`SSEC_PASSTHROUGH_CAPABILITY_TTL`], after which the next attempt +/// re-audits. Re-exported so existing `bucket_target_sys` consumers keep +/// their import path while the verdict vocabulary lives with the +/// replication decision logic. +pub use crate::bucket::replication::SsecPassthroughCapability; + +/// How long an audited SSE-C passthrough verdict stays authoritative. +/// +/// Trade-off: without a TTL a verdict is sticky for the process lifetime — +/// an `Unsupported` target that gets upgraded (or re-probed only via +/// replication-check) would keep failing SSE-C replication forever, and the +/// fail-open twin: a `Supported` verdict would outlive a backend swapped +/// behind the same endpoint/ARN. With the TTL, a bad target costs at most +/// one wasted PUT+HEAD audit per TTL window, and a changed backend is +/// re-discovered within the same window. +pub const SSEC_PASSTHROUGH_CAPABILITY_TTL: Duration = Duration::from_secs(10 * 60); + +/// A recorded SSE-C passthrough verdict plus when it was recorded, so reads +/// can report staleness against [`SSEC_PASSTHROUGH_CAPABILITY_TTL`]. +#[derive(Debug, Clone, Copy)] +struct SsecPassthroughRecord { + capability: SsecPassthroughCapability, + recorded_at: Instant, +} + #[derive(Debug, Default)] pub struct BucketTargetSys { pub arn_remotes_map: Arc>>, + /// SSE-C passthrough capability verdicts keyed by target ARN. See + /// [`SsecPassthroughCapability`]; reset alongside `arn_remotes_map`. + ssec_passthrough_map: Arc>>, pub targets_map: Arc>>>, pub h_mutex: Arc>>, target_h_mutex: Arc>>, @@ -317,6 +354,7 @@ impl BucketTargetSys { fn new() -> Self { Self { arn_remotes_map: Arc::new(RwLock::new(HashMap::new())), + ssec_passthrough_map: Arc::new(RwLock::new(HashMap::new())), targets_map: Arc::new(RwLock::new(HashMap::new())), h_mutex: Arc::new(RwLock::new(HashMap::new())), target_h_mutex: Arc::new(RwLock::new(HashMap::new())), @@ -580,19 +618,59 @@ impl BucketTargetSys { let update_mutex = self.target_update_mutex(bucket).await; let _update_guard = update_mutex.lock().await; - // Lock order: targets_map, then arn_remotes_map, then target_h_mutex. + // Lock order: targets_map, then arn_remotes_map, then target_h_mutex, + // then ssec_passthrough_map (always last; also taken standalone by the + // capability accessors). let mut targets_map = self.targets_map.write().await; let mut arn_remotes_map = self.arn_remotes_map.write().await; let mut health_map = self.target_h_mutex.write().await; if let Some(targets) = targets_map.remove(bucket) { + let mut ssec_map = self.ssec_passthrough_map.write().await; for target in targets { arn_remotes_map.remove(&target.arn); health_map.remove(&target.arn); + ssec_map.remove(&target.arn); } } } + /// Cached SSE-C passthrough capability for a target ARN, plus whether the + /// verdict is older than [`SSEC_PASSTHROUGH_CAPABILITY_TTL`]. `(Unknown, + /// false)` when no verdict has been recorded since the target was built. + /// Staleness is computed here so the gate policy stays a pure function. + pub async fn ssec_passthrough_capability(&self, arn: &str) -> (SsecPassthroughCapability, bool) { + match self.ssec_passthrough_map.read().await.get(arn) { + Some(record) => (record.capability, record.recorded_at.elapsed() >= SSEC_PASSTHROUGH_CAPABILITY_TTL), + None => (SsecPassthroughCapability::Unknown, false), + } + } + + /// Record an audited SSE-C passthrough verdict for a target ARN. Written by + /// the replication worker's HEAD-back audit and by the replication-check + /// SsecPassthrough probe phase. + pub async fn record_ssec_passthrough_capability(&self, arn: &str, capability: SsecPassthroughCapability) { + self.ssec_passthrough_map.write().await.insert( + arn.to_string(), + SsecPassthroughRecord { + capability, + recorded_at: Instant::now(), + }, + ); + } + + /// Test hook: age an existing verdict so TTL expiry is observable without + /// waiting out the real window. + #[cfg(test)] + pub(crate) async fn backdate_ssec_passthrough_capability(&self, arn: &str, age: Duration) { + let backdated = Instant::now() + .checked_sub(age) + .expect("system uptime must exceed the backdate age"); + if let Some(record) = self.ssec_passthrough_map.write().await.get_mut(arn) { + record.recorded_at = backdated; + } + } + pub async fn set_target( &self, bucket: &str, @@ -948,15 +1026,21 @@ impl BucketTargetSys { } } - // Lock order: targets_map, then arn_remotes_map, then target_h_mutex. + // Lock order: targets_map, then arn_remotes_map, then target_h_mutex, + // then ssec_passthrough_map (always last; also taken standalone by the + // capability accessors). let mut targets_map = self.targets_map.write().await; let mut arn_remotes_map = self.arn_remotes_map.write().await; let mut health_map = self.target_h_mutex.write().await; // Remove existing targets if let Some(existing_targets) = targets_map.remove(bucket) { + let mut ssec_map = self.ssec_passthrough_map.write().await; for target in existing_targets { arn_remotes_map.remove(&target.arn); health_map.remove(&target.arn); + // A rebuilt/edited target may point at a different service: + // the SSE-C passthrough verdict must be re-audited from Unknown. + ssec_map.remove(&target.arn); self.update_bandwidth_limit(bucket, &target.arn, 0); } } @@ -1446,6 +1530,43 @@ fn resolve_put_api_version_id(source_version_id: &str) -> Option<&str> { } } +/// Resolve the S3 `versionId` for a proxied read against a remote target. +/// RustFS represents the null version internally as the nil UUID while the S3 +/// API addresses it as the literal "null" (same mapping as +/// [`resolve_put_api_version_id`]); empty means "no version requested". +pub(crate) fn resolve_read_api_version_id(version_id: Option) -> Option { + let version_id = version_id?; + let trimmed = version_id.trim(); + if trimmed.is_empty() { + None + } else if Uuid::parse_str(trimmed).is_ok_and(|uuid| uuid.is_nil()) { + Some(rustfs_filemeta::NULL_VERSION_ID.to_string()) + } else { + Some(trimmed.to_string()) + } +} + +/// Outbound header set for a proxied read: the caller-provided passthrough +/// headers (client SSE-C key family, conditional headers) plus the anti-loop +/// `source-proxy-request` marker in both the x-rustfs- and x-minio- prefixes +/// (a MinIO target only understands the latter). Never adds +/// `source-replication-check`: that exemption channel belongs exclusively to +/// the replication worker's HEAD. +fn proxy_outbound_headers(mut extra_headers: HeaderMap) -> HeaderMap { + insert_header(&mut extra_headers, SUFFIX_SOURCE_PROXY_REQUEST, "true"); + extra_headers +} + +/// Copy `headers` onto an SDK request inside `customize().map_request` (runs +/// before signing, so the headers join the SigV4 canonical request). +fn apply_extra_headers(mut req: HttpRequest, headers: &HeaderMap) -> Result { + for (k, v) in headers.iter() { + req.headers_mut() + .insert(k.as_str().to_string(), v.to_str().unwrap_or("").to_string()); + } + Ok(req) +} + /// Append `versionId=` to an already-built request URI. aws-sdk-s3's /// `PutObjectInput` / `CreateMultipartUploadInput` expose no version id /// member, so the query is spliced in via `map_request`, which runs at @@ -1853,6 +1974,13 @@ impl TargetClient { // worker cannot hold; otherwise SSE-C replicas never converge on HEAD. let mut headers = HeaderMap::new(); insert_header(&mut headers, SUFFIX_SOURCE_REPLICATION_CHECK, "true"); + // `source-proxy-request: false` (MinIO `ProxyHeaderSet` semantics): + // the header's mere presence tells the receiver to answer LOCALLY + // instead of proxying the miss back to us. Without it, a not-found on + // the target gets read-proxied back to this source, echoes the source + // object with an identical ETag, and the worker concludes the object + // already converged — so it never actually replicates it. + insert_header(&mut headers, SUFFIX_SOURCE_PROXY_REQUEST, "false"); match self .client .head_object() @@ -1877,6 +2005,129 @@ impl TargetClient { } } + /// HEAD used by the read-proxy path (GET/HEAD of an object not yet + /// replicated locally, MinIO `proxyHeadToRepTarget`). + /// + /// Deliberately different from [`TargetClient::head_object`]: it must NOT + /// send `source-replication-check` — that header is the replication + /// worker's SSE-C metadata exemption channel. A proxied client request + /// instead forwards the client's own SSE-C headers (`extra_headers`) so + /// the target performs the real SSE-C validation/decryption. The + /// `source-proxy-request` marker is always added so the target does not + /// proxy the request onward (anti-loop). + pub async fn head_object_for_proxy( + &self, + bucket: &str, + object: &str, + version_id: Option, + range: Option, + part_number: Option, + extra_headers: HeaderMap, + ) -> Result> { + let headers = proxy_outbound_headers(extra_headers); + self.client + .head_object() + .bucket(bucket) + .key(object) + .set_version_id(resolve_read_api_version_id(version_id)) + .set_range(range) + .set_part_number(part_number) + .customize() + .map_request(move |req| apply_extra_headers(req, &headers)) + .send() + .await + } + + /// GET used by the read-proxy path (MinIO `proxyGetToReplicationTarget`). + /// Returns the streaming SDK output; callers must forward the body without + /// buffering it. Same header contract as [`Self::head_object_for_proxy`]: + /// anti-loop marker on, replication-check never sent, client SSE-C / + /// conditional headers forwarded verbatim via `extra_headers`. + pub async fn get_object( + &self, + bucket: &str, + object: &str, + version_id: Option, + range: Option, + part_number: Option, + extra_headers: HeaderMap, + ) -> Result> { + let headers = proxy_outbound_headers(extra_headers); + self.client + .get_object() + .bucket(bucket) + .key(object) + .set_version_id(resolve_read_api_version_id(version_id)) + .set_range(range) + .set_part_number(part_number) + .customize() + .map_request(move |req| apply_extra_headers(req, &headers)) + .send() + .await + } + + /// GetObjectTagging for the tagging read-proxy path + /// (MinIO `proxyGetTaggingToRepTarget`). Anti-loop marker always added. + pub async fn get_object_tagging( + &self, + bucket: &str, + object: &str, + version_id: Option, + ) -> Result> { + let headers = proxy_outbound_headers(HeaderMap::new()); + self.client + .get_object_tagging() + .bucket(bucket) + .key(object) + .set_version_id(resolve_read_api_version_id(version_id)) + .customize() + .map_request(move |req| apply_extra_headers(req, &headers)) + .send() + .await + } + + /// PutObjectTagging for the tagging proxy path + /// (MinIO `proxyTaggingToRepTarget`). Anti-loop marker always added. + pub async fn put_object_tagging( + &self, + bucket: &str, + object: &str, + version_id: Option, + tagging: SdkTagging, + ) -> Result> { + let headers = proxy_outbound_headers(HeaderMap::new()); + self.client + .put_object_tagging() + .bucket(bucket) + .key(object) + .set_version_id(resolve_read_api_version_id(version_id)) + .tagging(tagging) + .customize() + .map_request(move |req| apply_extra_headers(req, &headers)) + .send() + .await + } + + /// DeleteObjectTagging for the tagging proxy path + /// (MinIO `proxyTaggingToRepTarget`). Anti-loop marker always added. + pub async fn delete_object_tagging( + &self, + bucket: &str, + object: &str, + version_id: Option, + ) -> Result> { + let headers = proxy_outbound_headers(HeaderMap::new()); + self.client + .delete_object_tagging() + .bucket(bucket) + .key(object) + .set_version_id(resolve_read_api_version_id(version_id)) + .customize() + .map_request(move |req| apply_extra_headers(req, &headers)) + .send() + .await + } + /// On success returns the version id the target assigned (from /// `x-amz-version-id`), letting callers audit the version-identity /// contract — a target that adopts the source version echoes it back. @@ -2506,6 +2757,57 @@ mod tests { assert_eq!(health.last_online, Some(now)); } + /// N2 TTL contract, both flip directions: a recorded verdict is fresh + /// until [`SSEC_PASSTHROUGH_CAPABILITY_TTL`], then reads as expired; a + /// re-audit that records the OPPOSITE verdict replaces it as fresh. The + /// worker gate maps expired verdicts to ProceedWithAudit (pinned in + /// `replication_target_boundary`), so together this proves an Unsupported + /// target recovers to Supported through the audit once its verdict ages + /// out — and a stale Supported one is re-proven rather than trusted. + #[tokio::test] + async fn ssec_passthrough_capability_ttl_expires_and_reaudit_flips_verdict() { + let sys = BucketTargetSys::default(); + let arn = "arn:rustfs:replication:us-east-1:bucket:ssec-ttl"; + let expired_age = SSEC_PASSTHROUGH_CAPABILITY_TTL + Duration::from_secs(1); + + assert_eq!( + sys.ssec_passthrough_capability(arn).await, + (SsecPassthroughCapability::Unknown, false), + "an unrecorded target must read Unknown and never expired" + ); + + sys.record_ssec_passthrough_capability(arn, SsecPassthroughCapability::Unsupported) + .await; + assert_eq!( + sys.ssec_passthrough_capability(arn).await, + (SsecPassthroughCapability::Unsupported, false) + ); + + sys.backdate_ssec_passthrough_capability(arn, expired_age).await; + assert_eq!( + sys.ssec_passthrough_capability(arn).await, + (SsecPassthroughCapability::Unsupported, true), + "an aged-out Unsupported verdict must read expired so the gate re-audits" + ); + + // The re-audit against an upgraded target records Supported afresh. + sys.record_ssec_passthrough_capability(arn, SsecPassthroughCapability::Supported) + .await; + assert_eq!( + sys.ssec_passthrough_capability(arn).await, + (SsecPassthroughCapability::Supported, false), + "a fresh Supported verdict replaces the expired Unsupported one" + ); + + // And the fail-open twin: Supported also ages out. + sys.backdate_ssec_passthrough_capability(arn, expired_age).await; + assert_eq!( + sys.ssec_passthrough_capability(arn).await, + (SsecPassthroughCapability::Supported, true), + "an aged-out Supported verdict must read expired so the gate re-proves it" + ); + } + #[tokio::test] async fn list_targets_applies_health_stats_by_arn_and_preserves_endpoint_port() { let sys = BucketTargetSys::default(); diff --git a/crates/ecstore/src/bucket/replication/README.md b/crates/ecstore/src/bucket/replication/README.md index 68f43a32b..17e9ed7f6 100644 --- a/crates/ecstore/src/bucket/replication/README.md +++ b/crates/ecstore/src/bucket/replication/README.md @@ -11,9 +11,9 @@ paths. | Module | Current role | Split blocker | |---|---|---| | `config.rs` | Replication config helpers, rule matching, and tag filtering. | Uses replication-local filemeta/tagging boundaries and S3 DTOs directly. | -| `datatypes.rs` | ECStore compatibility re-export for resync status enums. | Re-exports `rustfs-replication` contracts while downstream facade consumers migrate. | | `replication_object_decision_boundary.rs` | Object replication option DTOs, resync target projection, delete replication decisions, and multipart planning helpers. | Keeps ECStore runtime modules from importing object decision contracts directly from `rustfs-replication`. | | `replication_pool.rs` | Replication queue, worker pool, MRF persistence, bucket stats, and delete/object scheduling. | Depends on bucket target sys, bucket metadata sys, metadata paths, queue contracts through the queue boundary, file metadata replication contracts through local boundaries, config storage, storage contracts through the replication storage boundary, runtime sources, and notification state. | +| `replication_proxy.rs` | Proxy-target selection for GET/HEAD/Tagging reads of objects not yet replicated locally (MinIO `getProxyTargets` parity: anti-loop, version-suspended, and no-config empty branches). | Uses replication config lookup, rule matching, and target clients through local boundaries. | | `replication_queue_boundary.rs` | Queue/admission DTOs, heal queue DTOs, worker sizing, and backpressure helpers. | Keeps ECStore runtime modules from importing queue/backpressure contracts directly from `rustfs-replication`. | | `replication_resync_boundary.rs` | Resync DTOs, status classifiers, persisted resync/MRF codec wrappers, and ECStore error mapping. | Keeps ECStore runtime modules from importing resync contract helpers directly from `rustfs-replication`. | | `replication_resyncer.rs` | Object replication, delete replication, resync execution, target calls, and multipart target upload paths. | Depends on target calls and target config types through the replication target boundary, metadata paths and metadata systems through the replication metadata boundary, file metadata replication contracts through the filemeta boundary, object decisions and multipart planning through the object decision boundary, resync contracts through the resync boundary, queue DTOs through the queue boundary, error contracts through the error boundary, versioning systems, storage contracts through the replication storage boundary, config-derived storage class labels through the config store, runtime sources, notification events and local event host selection through the event sink, bandwidth reader wrapping, and SetDisks lock timing. | @@ -117,9 +117,12 @@ Target end state: their file names — so batch-merging them beforehand is explicitly rejected: it forces synchronized guard-script/mod/import churn with zero functional gain; -- the only module that can retire early is `datatypes.rs`: delete it once its - facade consumers import the resync status enums through `rustfs-replication` - directly. +- `datatypes.rs` retired early (its sanctioned exception): it was a pure + relay (`boundary -> datatypes -> mod.rs`), so the facade now re-exports + `ResyncStatusType` from the resync boundary directly and the relay file is + deleted. Note the original retirement wording ("consumers import through + `rustfs-replication` directly") conflicted with Migration Rule #15 — + consumers stay behind the ECStore facade; only the relay hop dissolves. ## Milestones @@ -127,9 +130,9 @@ Target end state: |---|---|---| | M0 | Record the completion criteria and end state (this section). | Done | | M1 | Contract extraction: resync/queue/stats/object-decision/filemeta/storage wire contracts owned by `crates/replication`; ECStore imports concentrated in `*_boundary.rs`; event sink and runtime access behind local contracts. | Done — see Required Contracts | -| M2 | Move resyncer pure decision logic (no IO) into `crates/replication`. | Pending; sequence after splitting the oversized resyncer/pool functions (`resync_bucket`, `replicate_all`, `start_mrf_processor`) so moves stay mechanical | +| M2 | Move resyncer pure decision logic (no IO) into `crates/replication`. | Done — moved the pure decision helpers with their unit tests: `resync_status_duration` (resync), `resync_existing_delete_replication_info` / `replicate_delete_outcome` / `target_delete_version_id` / `delete_marker_purge_version_id` / `delete_marker_purge_mrf_entry` (delete), `version_identity_drifted` / `is_replication_target_offline_error` / the SSE-C passthrough gate family incl. `SsecPassthroughCapability` (object; `ssec_passthrough_evidence_present` was param-demoted to the echoed customer-algorithm string, ECStore keeps the `HeadObjectOutput` adapter). ECStore imports them through the resync/object-decision/target boundaries; `bucket_target_sys` keeps only the verdict cache + TTL and re-exports the capability enum. Not moved (signatures carry ECStore or aws-sdk types): `verify_resync_head_result`, `resync_target_error_detail`, the `SdkError` classifiers (`has_raw_status`, `is_version_id_format_mismatch`), the `replicate_all_*` option/info builders, and `bounded_resync_max_jobs` (itself a pure clamp, but it forms one local configuration unit with the env-reading `configured_resync_max_jobs` and its ECStore-local constants — moving the clamp alone has negative value). | | M3 | Move the worker runtime (`replication_pool.rs`, the IO paths of `replication_resyncer.rs`, `replication_state.rs`) once the contract traits are stable. Highest-risk step of the whole plan; do it last. | Pending | -| M4 | Retire the boundary modules together with their guard-script entries; delete `datatypes.rs`. | Pending | +| M4 | Retire the boundary modules together with their guard-script entries. | Pending (`datatypes.rs` already retired early alongside M2) | The original first code-bearing step (narrow `ReplicationEventSink` / `ReplicationRuntime` contracts) has landed — `replication_event_sink.rs` diff --git a/crates/ecstore/src/bucket/replication/datatypes.rs b/crates/ecstore/src/bucket/replication/datatypes.rs deleted file mode 100644 index 005a47966..000000000 --- a/crates/ecstore/src/bucket/replication/datatypes.rs +++ /dev/null @@ -1,15 +0,0 @@ -// Copyright 2024 RustFS Team -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// http://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -pub use super::replication_resync_boundary::ResyncStatusType; diff --git a/crates/ecstore/src/bucket/replication/mod.rs b/crates/ecstore/src/bucket/replication/mod.rs index d46f78eb3..be0de0121 100644 --- a/crates/ecstore/src/bucket/replication/mod.rs +++ b/crates/ecstore/src/bucket/replication/mod.rs @@ -12,7 +12,6 @@ // See the License for the specific language governing permissions and // limitations under the License. -pub mod datatypes; mod replication_bandwidth_boundary; mod replication_config_boundary; mod replication_config_store; @@ -29,6 +28,7 @@ mod replication_object_bridge; mod replication_object_config; mod replication_object_decision_boundary; pub(crate) mod replication_pool; +mod replication_proxy; mod replication_queue_boundary; mod replication_resync_boundary; mod replication_resyncer; @@ -43,7 +43,6 @@ pub(crate) mod replication_timing; mod replication_versioning_boundary; mod runtime_boundary; -pub use datatypes::ResyncStatusType; pub use replication_config_boundary::{ ObjectOpts, REMOTE_TARGET_CAPABILITY_CONTRACT_VERSION, REMOTE_TARGET_UNSUPPORTED_FIELDS, REMOTE_TARGET_WRITABLE_FIELDS, REPLICATION_CAPABILITY_CONTRACT_VERSION, REPLICATION_READ_ONLY_HISTORICAL_FIELDS, REPLICATION_WRITABLE_FIELDS, @@ -74,13 +73,16 @@ pub use replication_pool::{ get_global_replication_pool, get_global_replication_stats, init_background_replication, persist_force_delete_intent, read_durable_mrf_backlog, resync_start_conflict_id, }; +pub use replication_proxy::get_proxy_targets; pub use replication_queue_boundary::{ DeletedObjectReplicationInfo, ReplicationBatchAdmission, ReplicationHealQueueResult, ReplicationOperation, ReplicationPriority, ReplicationQueueAdmission, }; +pub use replication_resync_boundary::ResyncStatusType; pub use replication_resync_boundary::{BucketReplicationResyncStatus, ResyncOpts, TargetReplicationResyncStatus}; pub use replication_scanner_bridge::ReplicationScannerBridge; pub use replication_state::{ReplicationStats, RuntimeReplicationTargetBacklog}; pub use replication_stats_boundary::{BucketReplicationStat, BucketReplicationStats, BucketStats, InQueueMetric, XferStats}; pub use replication_storage_boundary::{ReplicationObjectIO, ReplicationStorage}; +pub use replication_target_boundary::SsecPassthroughCapability; pub(crate) use replication_target_config_bridge::ReplicationTargetConfigBridge; diff --git a/crates/ecstore/src/bucket/replication/replication_filemeta_boundary.rs b/crates/ecstore/src/bucket/replication/replication_filemeta_boundary.rs index 5a89e0fb2..917ac65a8 100644 --- a/crates/ecstore/src/bucket/replication/replication_filemeta_boundary.rs +++ b/crates/ecstore/src/bucket/replication/replication_filemeta_boundary.rs @@ -12,12 +12,11 @@ // See the License for the specific language governing permissions and // limitations under the License. -pub(crate) use rustfs_filemeta::NULL_VERSION_ID; pub use rustfs_replication::{MrfOpKind, MrfReplicateEntry}; pub(crate) use rustfs_replication::{ - REPLICATE_EXISTING, REPLICATE_EXISTING_DELETE, REPLICATE_HEAL_DELETE, ReplicateTargetDecision, ReplicatedInfos, - ReplicatedTargetInfo, ReplicationAction, ReplicationWorkerOperation, ResyncDecision, get_replication_state, - parse_replicate_decision, replicate_decision_for_admitted_targets, target_reset_header, version_purge_statuses_map, + REPLICATE_EXISTING, REPLICATE_HEAL_DELETE, ReplicateTargetDecision, ReplicatedInfos, ReplicatedTargetInfo, ReplicationAction, + ReplicationWorkerOperation, ResyncDecision, get_replication_state, parse_replicate_decision, + replicate_decision_for_admitted_targets, target_reset_header, version_purge_statuses_map, }; pub use rustfs_replication::{ REPLICATE_INCOMING_DELETE, ReplicateDecision, ReplicateObjectInfo, ReplicationState, ReplicationStatusType, ReplicationType, diff --git a/crates/ecstore/src/bucket/replication/replication_object_decision_boundary.rs b/crates/ecstore/src/bucket/replication/replication_object_decision_boundary.rs index d6c4f1ceb..4360ed4ef 100644 --- a/crates/ecstore/src/bucket/replication/replication_object_decision_boundary.rs +++ b/crates/ecstore/src/bucket/replication/replication_object_decision_boundary.rs @@ -18,9 +18,10 @@ pub use rustfs_replication::{ should_use_existing_delete_replication_source, }; pub(crate) use rustfs_replication::{ - ReplicationDeleteSource, ReplicationMultipartPartInput, ReplicationResyncTargetObject, - delete_replication_missing_source_decision, delete_replication_object_opts, heal_uses_delete_replication_path, - is_retryable_delete_replication_head_error, is_version_delete_replication, replication_etags_match, - replication_multipart_complete_actual_size, replication_multipart_part_plan, resync_target_for_object, - should_retry_delete_marker_purge, + ReplicationDeleteSource, ReplicationMultipartPartInput, ReplicationResyncTargetObject, delete_marker_purge_mrf_entry, + delete_marker_purge_version_id, delete_replication_missing_source_decision, delete_replication_object_opts, + heal_uses_delete_replication_path, is_retryable_delete_replication_head_error, is_version_delete_replication, + replicate_delete_outcome, replication_etags_match, replication_multipart_complete_actual_size, + replication_multipart_part_plan, resync_existing_delete_replication_info, resync_target_for_object, + should_retry_delete_marker_purge, target_delete_version_id, }; diff --git a/crates/ecstore/src/bucket/replication/replication_pool.rs b/crates/ecstore/src/bucket/replication/replication_pool.rs index 58b3efdca..efe517b99 100644 --- a/crates/ecstore/src/bucket/replication/replication_pool.rs +++ b/crates/ecstore/src/bucket/replication/replication_pool.rs @@ -667,6 +667,368 @@ async fn acknowledge_mrf_recovery( Err(EcstoreError::PreconditionFailed) } +/// Acquires the MRF recovery leader lock for the startup replay. +/// Returns `None` (after logging) when the lock cannot be created or another +/// node is already processing the backlog. +async fn acquire_mrf_recovery_guard(storage: &Arc) -> Option { + let recovery_lock = match storage + .new_ns_lock( + ReplicationMetadataStore::rustfs_meta_bucket(), + ReplicationMetadataStore::MRF_REPLICATION_RECOVERY_LOCK, + ) + .await + { + Ok(lock) => lock, + Err(error) => { + warn!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + error = %error, + "Failed to create the MRF recovery leader lock" + ); + return None; + } + }; + match recovery_lock + .get_write_lock_quiet(ReplicationLockTiming::acquire_timeout()) + .await + { + Ok(guard) => Some(guard), + Err(_) => { + debug!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + "Another node is already processing the MRF recovery backlog" + ); + None + } + } +} + +/// Reads and decodes the on-disk MRF recovery file. +/// Returns `None` when there is nothing to replay: missing file (publishes an +/// empty available summary), read failure, or corrupt data (quarantined). +async fn load_mrf_recovery_entries(storage: &Arc) -> Option> { + let data = match ReplicationConfigStore::read(storage.clone(), ReplicationMetadataStore::MRF_REPLICATION_FILE).await { + Ok(d) => d, + Err(EcstoreError::ConfigNotFound) => { + set_durable_mrf_backlog_summary(DurableMrfBacklogSummary { + available: true, + buckets: Vec::new(), + }); + return None; + } + Err(e) => { + warn!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + error = %e, + "Failed to load MRF recovery file" + ); + return None; + } + }; + + match decode_mrf_file(&data) { + Ok(v) => Some(v), + Err(e) => { + warn!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + error = %e, + "Failed to decode MRF recovery file — preserving corrupt data" + ); + quarantine_mrf_file(storage, &data).await; + None + } + } +} + +/// Replays one MRF recovery entry by operation kind. +/// Returns `None` when the entry is skipped entirely (no admission outcome); +/// entries that must be retried later are pushed onto `retry_entries`. +async fn replay_mrf_entry( + entry: &MrfReplicateEntry, + storage: &Arc, + retry_entries: &mut Vec, +) -> Option { + match entry.op { + MrfOpKind::Delete => replay_mrf_delete_entry(entry, storage, retry_entries).await, + MrfOpKind::Object | MrfOpKind::Heal | MrfOpKind::ExistingObject => { + replay_mrf_object_entry(entry, storage, retry_entries).await + } + MrfOpKind::Metadata => replay_mrf_metadata_entry(entry, storage, retry_entries).await, + } +} + +/// Replays a delete-kind MRF entry: force-delete intents replay directly, +/// stale force-delete generations are skipped, and plain deletes are +/// reconstructed as heal deletes. +async fn replay_mrf_delete_entry( + entry: &MrfReplicateEntry, + storage: &Arc, + retry_entries: &mut Vec, +) -> Option { + if should_replay_force_delete_intent(entry) { + let operation_id = entry.force_delete_id?; + let delete = force_delete_heal_replication_info(entry, operation_id); + if replicate_delete_with_outcome(delete, storage.clone()).await { + Some(ReplicationQueueAdmission::Queued) + } else { + Some(ReplicationQueueAdmission::Missed) + } + } else if entry.force_delete_id.is_some() { + Some(ReplicationQueueAdmission::Skipped) + } else { + replay_mrf_reconstructed_delete(entry, storage, retry_entries).await + } +} + +/// Pure DTO construction: heal replication info for a replayed force-delete intent. +fn force_delete_heal_replication_info(entry: &MrfReplicateEntry, operation_id: uuid::Uuid) -> DeletedObjectReplicationInfo { + DeletedObjectReplicationInfo { + delete_object: ReplicationDeletedObject { + object_name: entry.object.clone(), + force_delete: true, + force_delete_id: Some(operation_id), + force_delete_target_arns: entry.target_arns.clone(), + force_delete_generation: entry.force_delete_generation, + ..Default::default() + }, + bucket: entry.bucket.clone(), + op_type: ReplicationType::Heal, + event_type: REPLICATE_HEAL_DELETE.to_string(), + ..Default::default() + } +} + +/// Reconstruct a heal delete and re-queue it. We do NOT call +/// get_object_info here because the delete-marker or version may +/// already be absent from the local store — that is expected. +async fn replay_mrf_reconstructed_delete( + entry: &MrfReplicateEntry, + storage: &Arc, + retry_entries: &mut Vec, +) -> Option { + let versioned = ReplicationVersioningStore::prefix_enabled(&entry.bucket, &entry.object).await; + let oi = ObjectInfo { + bucket: entry.bucket.clone(), + name: entry.object.clone(), + version_id: entry.version_id, + delete_marker: entry.delete_marker, + ..Default::default() + }; + let dsc = resolve_mrf_delete_replicate_decision(entry, &oi, versioned, retry_entries).await?; + let dv = reconstructed_heal_delete_info(entry, &oi, &dsc); + if replicate_delete_with_outcome(dv, storage.clone()).await { + Some(ReplicationQueueAdmission::Queued) + } else { + Some(ReplicationQueueAdmission::Missed) + } +} + +/// The MRF entry does not persist the replication decision and the +/// source object is gone, so re-derive the decision from the live +/// bucket config (mirroring get_heal_replicate_object_info) and set +/// it on the reconstructed delete. Without this the decision string +/// is empty and the delete replicates to zero targets — a silent +/// no-op that leaves replicas diverged (backlog#858 / #799 B9). +async fn resolve_mrf_delete_replicate_decision( + entry: &MrfReplicateEntry, + oi: &ObjectInfo, + versioned: bool, + retry_entries: &mut Vec, +) -> Option { + if entry.target_arns.is_empty() { + match ReplicationMetadataStore::optional_replication_config(&entry.bucket).await { + Ok(None) => None, + Err(_) => { + retry_entries.push(entry.clone()); + None + } + Ok(Some(_)) => match check_replicate_delete_strict( + &entry.bucket, + &ObjectToDelete { + object_name: entry.object.clone(), + version_id: entry.version_id, + ..Default::default() + }, + oi, + &ObjectOptions { + versioned, + ..Default::default() + }, + None, + ) + .await + { + Ok(dsc) => Some(dsc), + Err(_) => { + retry_entries.push(entry.clone()); + None + } + }, + } + } else { + Some(replicate_decision_for_admitted_targets(&entry.target_arns)) + } +} + +/// Pure DTO construction: reconstructed heal delete carrying the re-derived +/// replication decision. +fn reconstructed_heal_delete_info( + entry: &MrfReplicateEntry, + oi: &ObjectInfo, + dsc: &ReplicateDecision, +) -> DeletedObjectReplicationInfo { + let mut rstate = oi.replication_state(); + rstate.replicate_decision_str = dsc.to_string(); + + let delete_marker_mtime = entry + .delete_marker_mtime + .and_then(|nanos| OffsetDateTime::from_unix_timestamp_nanos(i128::from(nanos)).ok()); + + DeletedObjectReplicationInfo { + delete_object: ReplicationDeletedObject { + object_name: entry.object.clone(), + version_id: entry.version_id, + delete_marker_version_id: entry.delete_marker_version_id, + delete_marker: entry.delete_marker, + delete_marker_mtime, + force_delete: entry.force_delete, + replication_state: Some(rstate), + ..Default::default() + }, + bucket: entry.bucket.clone(), + op_type: ReplicationType::Heal, + event_type: REPLICATE_HEAL_DELETE.to_string(), + ..Default::default() + } +} + +/// Replays an Object/Heal/ExistingObject MRF entry against the live source object. +async fn replay_mrf_object_entry( + entry: &MrfReplicateEntry, + storage: &Arc, + retry_entries: &mut Vec, +) -> Option { + let opts = ObjectOptions { + version_id: entry.version_id.map(|u| u.to_string()), + ..Default::default() + }; + let oi = match storage.get_object_info(&entry.bucket, &entry.object, &opts).await { + Ok(oi) => oi, + Err(e) => { + debug!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + bucket = %entry.bucket, + object = %entry.object, + error = %e, + "MRF recovery: source object lookup failed" + ); + if should_retry_mrf_source_lookup(&e) { + retry_entries.push(entry.clone()); + } + return None; + } + }; + if entry.target_arns.is_empty() { + // Legacy entries predate target admission persistence. They cannot + // be safely attributed, so retain the old live-config fallback. + Some(queue_replication_heal(&entry.bucket, oi, entry.retry_count.max(0) as u32).await) + } else { + let roi = admitted_mrf_replicate_object(oi, entry, entry.op.replication_type()); + if replicate_object_with_outcome(roi, storage.clone()).await.1 { + Some(ReplicationQueueAdmission::Queued) + } else { + Some(ReplicationQueueAdmission::Missed) + } + } +} + +/// Replays a metadata-kind MRF entry against the live source object. +async fn replay_mrf_metadata_entry( + entry: &MrfReplicateEntry, + storage: &Arc, + retry_entries: &mut Vec, +) -> Option { + let opts = ObjectOptions { + version_id: entry.version_id.map(|u| u.to_string()), + ..Default::default() + }; + let oi = match storage.get_object_info(&entry.bucket, &entry.object, &opts).await { + Ok(oi) => oi, + Err(e) => { + debug!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + bucket = %entry.bucket, + object = %entry.object, + error = %e, + "MRF metadata recovery: source object lookup failed" + ); + if should_retry_mrf_source_lookup(&e) { + retry_entries.push(entry.clone()); + } + return None; + } + }; + if entry.target_arns.is_empty() { + Some(queue_replication_metadata(&entry.bucket, oi, entry.retry_count.max(0) as u32).await) + } else { + let roi = admitted_mrf_replicate_object(oi, entry, ReplicationType::Metadata); + if replicate_object_with_outcome(roi, storage.clone()).await.1 { + Some(ReplicationQueueAdmission::Queued) + } else { + Some(ReplicationQueueAdmission::Missed) + } + } +} + +/// Pure DTO construction: replicate-object info for an entry with persisted +/// admitted targets, carrying over the entry's retry count. +fn admitted_mrf_replicate_object(oi: ObjectInfo, entry: &MrfReplicateEntry, op_type: ReplicationType) -> ReplicateObjectInfo { + let dsc = replicate_decision_for_admitted_targets(&entry.target_arns); + let mut roi = replicate_object_info_from_object_info(oi, dsc, op_type); + roi.retry_count = entry.retry_count.max(0) as u32; + roi +} + +/// Acknowledges the replayed MRF prefix and returns the retained backlog. +/// On acknowledgement failure the backlog is preserved for the next startup and +/// re-read (falling back to the replayed snapshot) so the published summary stays accurate. +async fn resolve_retained_mrf_entries( + storage: &Arc, + recovery_guard: &rustfs_lock::NamespaceLockGuard, + entries: &[MrfReplicateEntry], + retry_entries: &[MrfReplicateEntry], +) -> Vec { + match acknowledge_mrf_recovery(storage.clone(), recovery_guard, entries, retry_entries).await { + Ok(retained) => retained, + Err(error) => { + warn!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + error = %error, + "Failed to acknowledge the MRF recovery prefix; preserving it for the next startup" + ); + match read_mrf_entries(storage.clone()).await { + Ok(current) => current, + Err(read_error) => { + warn!( + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION, + error = %read_error, + "Failed to refresh the MRF backlog after acknowledgement failure" + ); + entries.to_vec() + } + } + } + } +} + #[derive(Debug, thiserror::Error)] #[error("replication resync {active_resync_id} is already active for {bucket}/{arn}")] struct ResyncActiveConflictError { @@ -1221,71 +1583,12 @@ impl ReplicationPool { let storage = self.storage.clone(); let handle = tokio::spawn(async move { - let recovery_lock = match storage - .new_ns_lock( - ReplicationMetadataStore::rustfs_meta_bucket(), - ReplicationMetadataStore::MRF_REPLICATION_RECOVERY_LOCK, - ) - .await - { - Ok(lock) => lock, - Err(error) => { - warn!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - error = %error, - "Failed to create the MRF recovery leader lock" - ); - return; - } - }; - let recovery_guard = match recovery_lock - .get_write_lock_quiet(ReplicationLockTiming::acquire_timeout()) - .await - { - Ok(guard) => guard, - Err(_) => { - debug!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - "Another node is already processing the MRF recovery backlog" - ); - return; - } + let Some(recovery_guard) = acquire_mrf_recovery_guard(&storage).await else { + return; }; - let data = match ReplicationConfigStore::read(storage.clone(), ReplicationMetadataStore::MRF_REPLICATION_FILE).await { - Ok(d) => d, - Err(EcstoreError::ConfigNotFound) => { - set_durable_mrf_backlog_summary(DurableMrfBacklogSummary { - available: true, - buckets: Vec::new(), - }); - return; - } - Err(e) => { - warn!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - error = %e, - "Failed to load MRF recovery file" - ); - return; - } - }; - - let entries = match decode_mrf_file(&data) { - Ok(v) => v, - Err(e) => { - warn!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - error = %e, - "Failed to decode MRF recovery file — preserving corrupt data" - ); - quarantine_mrf_file(&storage, &data).await; - return; - } + let Some(entries) = load_mrf_recovery_entries(&storage).await else { + return; }; set_durable_mrf_backlog_snapshot(durable_mrf_backlog_summary_from_entries(&entries)); @@ -1294,187 +1597,8 @@ impl ReplicationPool { let mut retry_entries = Vec::new(); for entry in entries.iter() { - let admission = match entry.op { - MrfOpKind::Delete => { - if should_replay_force_delete_intent(entry) { - let Some(operation_id) = entry.force_delete_id else { - continue; - }; - let delete = DeletedObjectReplicationInfo { - delete_object: ReplicationDeletedObject { - object_name: entry.object.clone(), - force_delete: true, - force_delete_id: Some(operation_id), - force_delete_target_arns: entry.target_arns.clone(), - force_delete_generation: entry.force_delete_generation, - ..Default::default() - }, - bucket: entry.bucket.clone(), - op_type: ReplicationType::Heal, - event_type: REPLICATE_HEAL_DELETE.to_string(), - ..Default::default() - }; - if replicate_delete_with_outcome(delete, storage.clone()).await { - ReplicationQueueAdmission::Queued - } else { - ReplicationQueueAdmission::Missed - } - } else if entry.force_delete_id.is_some() { - ReplicationQueueAdmission::Skipped - } else { - // Reconstruct a heal delete and re-queue it. We do NOT call - // get_object_info here because the delete-marker or version may - // already be absent from the local store — that is expected. - // - // The MRF entry does not persist the replication decision and the - // source object is gone, so re-derive the decision from the live - // bucket config (mirroring get_heal_replicate_object_info) and set - // it on the reconstructed delete. Without this the decision string - // is empty and the delete replicates to zero targets — a silent - // no-op that leaves replicas diverged (backlog#858 / #799 B9). - let versioned = ReplicationVersioningStore::prefix_enabled(&entry.bucket, &entry.object).await; - let oi = ObjectInfo { - bucket: entry.bucket.clone(), - name: entry.object.clone(), - version_id: entry.version_id, - delete_marker: entry.delete_marker, - ..Default::default() - }; - let dsc = if entry.target_arns.is_empty() { - match ReplicationMetadataStore::optional_replication_config(&entry.bucket).await { - Ok(None) => continue, - Err(_) => { - retry_entries.push(entry.clone()); - continue; - } - Ok(Some(_)) => match check_replicate_delete_strict( - &entry.bucket, - &ObjectToDelete { - object_name: entry.object.clone(), - version_id: entry.version_id, - ..Default::default() - }, - &oi, - &ObjectOptions { - versioned, - ..Default::default() - }, - None, - ) - .await - { - Ok(dsc) => dsc, - Err(_) => { - retry_entries.push(entry.clone()); - continue; - } - }, - } - } else { - replicate_decision_for_admitted_targets(&entry.target_arns) - }; - let mut rstate = oi.replication_state(); - rstate.replicate_decision_str = dsc.to_string(); - - let delete_marker_mtime = entry - .delete_marker_mtime - .and_then(|nanos| OffsetDateTime::from_unix_timestamp_nanos(i128::from(nanos)).ok()); - - let dv = DeletedObjectReplicationInfo { - delete_object: ReplicationDeletedObject { - object_name: entry.object.clone(), - version_id: entry.version_id, - delete_marker_version_id: entry.delete_marker_version_id, - delete_marker: entry.delete_marker, - delete_marker_mtime, - force_delete: entry.force_delete, - replication_state: Some(rstate), - ..Default::default() - }, - bucket: entry.bucket.clone(), - op_type: ReplicationType::Heal, - event_type: REPLICATE_HEAL_DELETE.to_string(), - ..Default::default() - }; - if replicate_delete_with_outcome(dv, storage.clone()).await { - ReplicationQueueAdmission::Queued - } else { - ReplicationQueueAdmission::Missed - } - } - } - MrfOpKind::Object | MrfOpKind::Heal | MrfOpKind::ExistingObject => { - let opts = ObjectOptions { - version_id: entry.version_id.map(|u| u.to_string()), - ..Default::default() - }; - let oi = match storage.get_object_info(&entry.bucket, &entry.object, &opts).await { - Ok(oi) => oi, - Err(e) => { - debug!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - bucket = %entry.bucket, - object = %entry.object, - error = %e, - "MRF recovery: source object lookup failed" - ); - if should_retry_mrf_source_lookup(&e) { - retry_entries.push(entry.clone()); - } - continue; - } - }; - if entry.target_arns.is_empty() { - // Legacy entries predate target admission persistence. They cannot - // be safely attributed, so retain the old live-config fallback. - queue_replication_heal(&entry.bucket, oi, entry.retry_count.max(0) as u32).await - } else { - let dsc = replicate_decision_for_admitted_targets(&entry.target_arns); - let mut roi = replicate_object_info_from_object_info(oi, dsc, entry.op.replication_type()); - roi.retry_count = entry.retry_count.max(0) as u32; - if replicate_object_with_outcome(roi, storage.clone()).await.1 { - ReplicationQueueAdmission::Queued - } else { - ReplicationQueueAdmission::Missed - } - } - } - MrfOpKind::Metadata => { - let opts = ObjectOptions { - version_id: entry.version_id.map(|u| u.to_string()), - ..Default::default() - }; - let oi = match storage.get_object_info(&entry.bucket, &entry.object, &opts).await { - Ok(oi) => oi, - Err(e) => { - debug!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - bucket = %entry.bucket, - object = %entry.object, - error = %e, - "MRF metadata recovery: source object lookup failed" - ); - if should_retry_mrf_source_lookup(&e) { - retry_entries.push(entry.clone()); - } - continue; - } - }; - if entry.target_arns.is_empty() { - queue_replication_metadata(&entry.bucket, oi, entry.retry_count.max(0) as u32).await - } else { - let dsc = replicate_decision_for_admitted_targets(&entry.target_arns); - let mut roi = replicate_object_info_from_object_info(oi, dsc, ReplicationType::Metadata); - roi.retry_count = entry.retry_count.max(0) as u32; - if replicate_object_with_outcome(roi, storage.clone()).await.1 { - ReplicationQueueAdmission::Queued - } else { - ReplicationQueueAdmission::Missed - } - } - } + let Some(admission) = replay_mrf_entry(entry, &storage, &mut retry_entries).await else { + continue; }; if admission == ReplicationQueueAdmission::Missed { @@ -1484,29 +1608,7 @@ impl ReplicationPool { } } - let retained = match acknowledge_mrf_recovery(storage.clone(), &recovery_guard, &entries, &retry_entries).await { - Ok(retained) => retained, - Err(error) => { - warn!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - error = %error, - "Failed to acknowledge the MRF recovery prefix; preserving it for the next startup" - ); - match read_mrf_entries(storage.clone()).await { - Ok(current) => current, - Err(read_error) => { - warn!( - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION, - error = %read_error, - "Failed to refresh the MRF backlog after acknowledgement failure" - ); - entries.clone() - } - } - } - }; + let retained = resolve_retained_mrf_entries(&storage, &recovery_guard, &entries, &retry_entries).await; let retained_count = retained.len(); set_durable_mrf_backlog_snapshot(durable_mrf_backlog_summary_from_entries(&retained)); diff --git a/crates/ecstore/src/bucket/replication/replication_proxy.rs b/crates/ecstore/src/bucket/replication/replication_proxy.rs new file mode 100644 index 000000000..d71895f16 --- /dev/null +++ b/crates/ecstore/src/bucket/replication/replication_proxy.rs @@ -0,0 +1,150 @@ +// Copyright 2024 RustFS Team +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +//! Proxy-target selection for reads of objects not yet replicated locally +//! (MinIO `getProxyTargets`, bucket-replication.go). +//! +//! During the active-active replication lag window a GET/HEAD/Tagging request +//! for an object the local site does not have yet may be served by proxying to +//! a replication target. This module only *selects* the candidate targets; the +//! request-path callers perform the remote calls and response translation. + +use std::sync::Arc; + +use tracing::debug; + +use super::replication_config_boundary::{ObjectOpts, ReplicationConfigurationExt as _}; +use super::replication_object_config::get_replication_config; +use super::replication_storage_boundary::ObjectOptions; +use super::replication_target_boundary::{ReplicationTargetStore, TargetClient}; + +/// Returns the replication-target clients eligible to serve a proxied read of +/// `bucket/object`, in rule order. Mirrors MinIO's `getProxyTargets`: +/// +/// - the `source-proxy-request` header family was present at all +/// (`opts.proxy_request` / `opts.proxy_header_set`, MinIO `ProxyRequest` / +/// `ProxyHeaderSet`) -> empty. "true" is the anti-loop marker of an +/// already-proxied client read; "false" is what a peer's replication +/// worker sends on convergence HEADs so the receiver answers locally — +/// proxying that miss back would echo the source object and fake +/// convergence, permanently skipping replication; +/// - the bucket's versioning is suspended for the object -> empty; +/// - no replication configuration / no matching rule -> empty; +/// - otherwise every distinct target ARN whose rules match the object, +/// resolved through the bucket target system, skipping targets that opted +/// out of proxying (`disable_proxy`). +pub async fn get_proxy_targets(bucket: &str, object: &str, opts: &ObjectOptions) -> Vec> { + if opts.proxy_request || opts.proxy_header_set { + return Vec::new(); + } + if opts.version_suspended { + return Vec::new(); + } + + let cfg = match get_replication_config(bucket).await { + Ok(Some(cfg)) => cfg, + Ok(None) => return Vec::new(), + Err(err) => { + debug!(bucket, object, error = %err, "read proxy: failed to load replication config; not proxying"); + return Vec::new(); + } + }; + + let arns = cfg.filter_target_arns(&ObjectOpts { + name: object.to_string(), + ..Default::default() + }); + + let mut targets = Vec::with_capacity(arns.len()); + for arn in arns { + let Some(client) = ReplicationTargetStore::remote_target_client(bucket, &arn).await else { + debug!(bucket, object, arn, "read proxy: no client for replication target ARN"); + continue; + }; + if client.disable_proxy { + continue; + } + targets.push(client); + } + + targets +} + +#[cfg(test)] +mod tests { + use super::*; + + fn opts() -> ObjectOptions { + ObjectOptions::default() + } + + /// Anti-loop: a request that was already proxied by a peer must never be + /// proxied onward, regardless of replication configuration. + #[tokio::test] + async fn proxy_request_yields_no_targets() { + let targets = get_proxy_targets( + "bucket", + "object", + &ObjectOptions { + proxy_request: true, + ..opts() + }, + ) + .await; + assert!(targets.is_empty()); + } + + /// MinIO `ProxyHeaderSet` parity: the header family being present at all + /// disables proxying, even with the value "false" — that is what a + /// peer's replication worker sends on convergence HEADs. + #[tokio::test] + async fn proxy_header_set_yields_no_targets() { + let targets = get_proxy_targets( + "bucket", + "object", + &ObjectOptions { + proxy_header_set: true, + proxy_request: false, + ..opts() + }, + ) + .await; + assert!(targets.is_empty()); + } + + /// Suspended versioning disables proxying (MinIO parity): the local null + /// version is authoritative and a remote read could resurrect data. + #[tokio::test] + async fn version_suspended_yields_no_targets() { + let targets = get_proxy_targets( + "bucket", + "object", + &ObjectOptions { + version_suspended: true, + ..opts() + }, + ) + .await; + assert!(targets.is_empty()); + } + + /// A bucket without replication configuration has nothing to proxy to. + /// (No metadata system is running in unit tests, so the config lookup + /// resolves to "no configuration" — the same empty-result contract.) + #[tokio::test] + async fn missing_replication_config_yields_no_targets() { + let targets = get_proxy_targets("bucket-without-replication", "object", &opts()).await; + assert!(targets.is_empty()); + } +} diff --git a/crates/ecstore/src/bucket/replication/replication_resync_boundary.rs b/crates/ecstore/src/bucket/replication/replication_resync_boundary.rs index 3b6dcdf5d..32e3f6c61 100644 --- a/crates/ecstore/src/bucket/replication/replication_resync_boundary.rs +++ b/crates/ecstore/src/bucket/replication/replication_resync_boundary.rs @@ -15,10 +15,15 @@ use super::replication_error_boundary::{Error, Result}; use super::replication_filemeta_boundary::MrfReplicateEntry; +/// Kept test-only: the runtime consumer was the worker HEAD's fake proxy +/// counting (removed in backlog#1675 P1-5); the resyncer tests still pin the +/// classifier's semantics for the real client read-proxy failure accounting. +#[cfg(test)] +pub(crate) use rustfs_replication::should_count_head_proxy_failure; pub use rustfs_replication::{BucketReplicationResyncStatus, ResyncOpts, ResyncStatusType, TargetReplicationResyncStatus}; pub(crate) use rustfs_replication::{ - is_version_id_mismatch, resync_state_accepts_update, sanitize_resync_error_detail, should_auto_resume_resync, - should_count_head_proxy_failure, + is_version_id_mismatch, resync_state_accepts_update, resync_status_duration, sanitize_resync_error_detail, + should_auto_resume_resync, }; #[allow( diff --git a/crates/ecstore/src/bucket/replication/replication_resyncer.rs b/crates/ecstore/src/bucket/replication/replication_resyncer.rs index 30198a75b..882ece91c 100644 --- a/crates/ecstore/src/bucket/replication/replication_resyncer.rs +++ b/crates/ecstore/src/bucket/replication/replication_resyncer.rs @@ -15,13 +15,12 @@ use super::replication_bandwidth_boundary; use super::replication_config_boundary::{ObjectOpts, ReplicationConfigurationExt as _}; use super::replication_config_store::ReplicationConfigStore; -use super::replication_error_boundary::{Result, is_err_object_not_found, is_err_version_not_found}; +use super::replication_error_boundary::{Error, Result, is_err_object_not_found, is_err_version_not_found}; use super::replication_event_sink::{EventArgs, send_event, send_local_event}; use super::replication_filemeta_boundary::{ - MrfReplicateEntry, NULL_VERSION_ID, REPLICATE_EXISTING, REPLICATE_EXISTING_DELETE, ReplicateDecision, ReplicateObjectInfo, - ReplicatedInfos, ReplicatedTargetInfo, ReplicationAction, ReplicationState, ReplicationStatusType, ReplicationType, - ReplicationWorkerOperation, VersionPurgeStatusType, get_replication_state, parse_replicate_decision, - replication_statuses_map, target_reset_header, version_purge_statuses_map, + REPLICATE_EXISTING, ReplicateDecision, ReplicateObjectInfo, ReplicatedInfos, ReplicatedTargetInfo, ReplicationAction, + ReplicationState, ReplicationStatusType, ReplicationType, VersionPurgeStatusType, get_replication_state, + parse_replicate_decision, replication_statuses_map, target_reset_header, version_purge_statuses_map, }; use super::replication_lock_boundary::ReplicationLockTiming; use super::replication_logging::{EVENT_RESYNC_CONFIG_LOOKUP_SKIPPED, LOG_COMPONENT_ECSTORE, LOG_SUBSYSTEM_REPLICATION_RESYNC}; @@ -30,27 +29,35 @@ use super::replication_metadata_boundary::ReplicationMetadataStore; use super::replication_msgp_boundary::ReplicationMsgpCodec; use super::replication_object_config::{ReplicationConfig, get_replication_config, must_replicate}; use super::replication_object_decision_boundary::{ - MustReplicateOptions, ReplicationMultipartPartInput, heal_uses_delete_replication_path, - is_retryable_delete_replication_head_error, is_version_delete_replication, replication_etags_match, - replication_multipart_complete_actual_size, replication_multipart_part_plan, should_retry_delete_marker_purge, + MustReplicateOptions, ReplicationMultipartPartInput, delete_marker_purge_mrf_entry, delete_marker_purge_version_id, + heal_uses_delete_replication_path, is_retryable_delete_replication_head_error, is_version_delete_replication, + replicate_delete_outcome, replication_etags_match, replication_multipart_complete_actual_size, + replication_multipart_part_plan, resync_existing_delete_replication_info, should_retry_delete_marker_purge, + target_delete_version_id, }; use super::replication_queue_boundary::{DeletedObjectReplicationInfo, ReplicationQueueAdmission}; use super::replication_resync_boundary::ResyncStatusType; +#[cfg(test)] +use super::replication_resync_boundary::should_count_head_proxy_failure; use super::replication_resync_boundary::{ BucketReplicationResyncStatus, ResyncOpts, TargetReplicationResyncStatus, encode_resync_file, is_version_id_mismatch, - resync_state_accepts_update, sanitize_resync_error_detail, should_count_head_proxy_failure, + resync_state_accepts_update, resync_status_duration, sanitize_resync_error_detail, }; #[cfg(test)] use super::replication_resync_boundary::{RESYNC_META_FORMAT, RESYNC_META_VERSION, WIRE_ZERO_TIME_UNIX, decode_resync_file}; +#[cfg(test)] +use super::replication_storage_boundary::ReplicationDeletedObject; use super::replication_storage_boundary::{ - AdvancedGetOptions, EcstoreObjectOperations, HTTPRangeSpec, ObjectInfo, ObjectOptions, ObjectToDelete, - ReplicationDeletedObject, ReplicationObjectIO, ReplicationStorage, StatObjectOptions, WalkOptions, + AdvancedGetOptions, EcstoreObjectOperations, GetObjectReader, HTTPRangeSpec, ObjectInfo, ObjectOptions, ObjectToDelete, + ReplicationObjectIO, ReplicationStorage, StatObjectOptions, StorageObjectInfoOrErr, WalkOptions, }; use super::replication_target_boundary::{ - PutObjectOptions, PutObjectPartOptions, ReplicationTargetStore, TargetClient, replication_action_for_target_head, - replication_complete_multipart_options, replication_delete_marker_purge_remove_options, replication_delete_remove_options, - replication_force_delete_remove_options, replication_object_is_ssec_encrypted, replication_put_object_header_size, - replication_put_object_options, replication_target_head_is_newer_null_version, + ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED, PutObjectOptions, PutObjectPartOptions, ReplicationTargetStore, + SsecPassthroughCapability, SsecPassthroughGate, TargetClient, is_replication_target_offline_error, + replication_action_for_target_head, replication_complete_multipart_options, replication_delete_marker_purge_remove_options, + replication_delete_remove_options, replication_force_delete_remove_options, replication_object_is_ssec_encrypted, + replication_put_object_header_size, replication_put_object_options, replication_target_head_is_newer_null_version, + resolve_read_api_version_id, ssec_passthrough_evidence_present, ssec_passthrough_gate, version_identity_drifted, }; use super::replication_versioning_boundary::ReplicationVersioningStore; use super::runtime_boundary as runtime_sources; @@ -106,21 +113,6 @@ const EVENT_DELETE_MARKER_PURGE_FAILED: &str = "replication_delete_marker_purge_ const EVENT_DELETE_MARKER_PURGE_MRF: &str = "replication_delete_marker_purge_mrf"; const METRIC_DELETE_MARKER_PURGE_TOTAL: &str = "rustfs_replication_delete_marker_purge_total"; const EVENT_REPLICATION_VERSION_IDENTITY_DRIFT: &str = "replication_version_identity_drift"; -const REPLICATION_TARGET_OFFLINE_ERROR_MARKERS: &[&str] = &[ - "dispatch failure", - "timeouterror", - "timed out", - "connection refused", - "connection reset", - "connection closed", - "connection aborted", - "broken pipe", - "dns error", - "failed to lookup address", - "name or service not known", - "deadline has elapsed", - "tcp connect error", -]; #[allow( dead_code, @@ -182,15 +174,6 @@ fn has_raw_status(err: &SdkError, status: u16) -> bool { err.raw_response().is_some_and(|r| r.status().as_u16() == status) } -fn is_head_proxy_failure(err: &SdkError) -> bool { - let (is_not_found, code) = err - .as_service_error() - .map(|service_err| (service_err.is_not_found(), service_err.code())) - .unwrap_or((false, None)); - let raw_status = err.raw_response().map(|resp| resp.status().as_u16()); - should_count_head_proxy_failure(is_not_found, code, raw_status) -} - const METRIC_VERSION_IDENTITY_DRIFT_TOTAL: &str = "rustfs_replication_version_identity_drift_total"; /// Targets that already produced a version-identity-drift warning this @@ -199,27 +182,6 @@ const METRIC_VERSION_IDENTITY_DRIFT_TOTAL: &str = "rustfs_replication_version_id /// after a restart is acceptable. static VERSION_IDENTITY_WARNED_ARNS: LazyLock>> = LazyLock::new(|| StdMutex::new(HashSet::new())); -/// Runtime half of the P1-19 version-identity contract (the explicit probe -/// lives in replication-check's VersionFidelity phase): every replication PUT -/// response reveals whether the target adopted the source version id. A -/// target minting its own ids silently breaks version-addressed deletes and -/// heal, so surface it — once per target — instead of letting the divergence -/// accumulate unseen. -/// Pure drift judgment: the contract only applies when the source addressed a -/// real (non-nil) version uuid, and drift means the target answered with -/// anything else — including nothing at all. -fn version_identity_drifted(source_version_id: &str, assigned_version_id: Option<&str>) -> bool { - if source_version_id.is_empty() { - return false; - } - // A nil source uuid travels as the literal "null" (unversioned-source - // semantics); no identity contract applies to it. - if Uuid::parse_str(source_version_id).map(|uuid| uuid.is_nil()).unwrap_or(true) { - return false; - } - assigned_version_id != Some(source_version_id) -} - fn audit_target_version_identity(tgt_client: &TargetClient, source_version_id: &str, assigned_version_id: Option<&str>) { if !version_identity_drifted(source_version_id, assigned_version_id) { return; @@ -242,23 +204,18 @@ fn audit_target_version_identity(tgt_client: &TargetClient, source_version_id: & } } -async fn record_proxy_request(bucket: &str, api: &str, is_err: bool) { - if let Some(stats) = runtime_sources::replication_stats() { - stats.inc_proxy(bucket, api, is_err).await; - } -} - -async fn head_object_with_proxy_stats( - source_bucket: &str, +/// HEAD against a replication target on behalf of the replication worker +/// (resync/heal/delete convergence checks). This is NOT a client read proxy: +/// it must not touch the proxy metrics — those count only real GET/HEAD/ +/// Tagging requests proxied for clients (see `replication_proxy.rs` / +/// `TargetClient::head_object_for_proxy`). +async fn head_object_for_worker( target_client: &TargetClient, target_bucket: &str, object: &str, version_id: Option, ) -> std::result::Result> { - let result = target_client.head_object(target_bucket, object, version_id).await; - let is_err = result.as_ref().err().is_some_and(is_head_proxy_failure); - record_proxy_request(source_bucket, "HeadObject", is_err).await; - result + target_client.head_object(target_bucket, object, version_id).await } fn is_version_id_format_mismatch(err: &SdkError) -> bool { @@ -267,13 +224,6 @@ fn is_version_id_format_mismatch(err: &SdkError) -> bool { is_version_id_mismatch(code, raw_status) } -fn is_replication_target_offline_error(err: &(impl Display + ?Sized)) -> bool { - let message = err.to_string().to_ascii_lowercase(); - REPLICATION_TARGET_OFFLINE_ERROR_MARKERS - .iter() - .any(|marker| message.contains(marker)) -} - async fn mark_replication_target_offline_if_needed(target_client: &Arc, err: &(impl Display + ?Sized)) { if is_replication_target_offline_error(err) { ReplicationTargetStore::mark_target_offline(target_client).await; @@ -281,44 +231,126 @@ async fn mark_replication_target_offline_if_needed(target_client: &Arc std::result::Result, SdkError> { - match head_object_with_proxy_stats(source_bucket, tgt_client, &tgt_client.bucket, object, None).await { + match head_object_for_worker(tgt_client, &tgt_client.bucket, object, None).await { Ok(oi) => Ok(Some(oi)), Err(e) if e.as_service_error().is_some_and(|se| se.is_not_found()) || has_raw_status(&e, 404) => Ok(None), Err(e) => Err(e), } } -static RESYNC_WORKER_COUNT: usize = 10; - -fn resync_status_duration( - status: ResyncStatusType, - start_time: Option, - now: OffsetDateTime, -) -> Option { - if !matches!( - status, - ResyncStatusType::ResyncCompleted | ResyncStatusType::ResyncFailed | ResyncStatusType::ResyncCanceled - ) { - return None; +/// Resolve the N2 fail-closed gate for an SSE-C passthrough attempt against +/// this target. Returns `Some(audit_required)` when replication may proceed; +/// on a freshly-flagged header-dropping target it settles `rinfo` as FAILED +/// (no PUT is ever sent — the object stays on the normal MRF retry channel +/// and re-audits once the verdict's TTL expires or replication-check +/// re-probes the target) and returns `None`. +async fn resolve_ssec_passthrough_gate( + ssec: bool, + tgt_client: &TargetClient, + bucket: &str, + object: &str, + rinfo: &mut ReplicatedTargetInfo, +) -> Option { + let (capability, expired) = ReplicationTargetStore::ssec_passthrough_capability(&tgt_client.arn).await; + match ssec_passthrough_gate(ssec, capability, expired) { + SsecPassthroughGate::Proceed => Some(false), + SsecPassthroughGate::ProceedWithAudit => Some(true), + SsecPassthroughGate::FailClosed => { + rinfo.replication_status = ReplicationStatusType::Failed; + rinfo.error = Some(ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED.to_string()); + warn!( + event = EVENT_RESYNC_TARGET_OPERATION_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + object = %object, + arn = %tgt_client.arn, + operation = "ssec_passthrough_gate", + error = ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED, + "Replication target operation failed" + ); + None + } } - - let millis = (now - start_time?).whole_milliseconds(); - if millis < 0 { - return None; - } - - let millis = if millis > i128::from(u64::MAX) { - u64::MAX - } else { - u64::try_from(millis).ok()? - }; - Some(std::time::Duration::from_millis(millis)) } +/// Judge SSE-C passthrough evidence on a HEAD of the replica and record the +/// capability verdict for the target. Returns true when the SSE-C material +/// provably survived; otherwise records `Unsupported` and settles `rinfo` as +/// FAILED so the attempt never reports a silently unreadable COMPLETED. +async fn settle_ssec_passthrough_evidence( + head: &HeadObjectOutput, + tgt_client: &TargetClient, + bucket: &str, + object: &str, + rinfo: &mut ReplicatedTargetInfo, +) -> bool { + if ssec_passthrough_evidence_present(head) { + ReplicationTargetStore::record_ssec_passthrough_capability(&tgt_client.arn, SsecPassthroughCapability::Supported).await; + return true; + } + ReplicationTargetStore::record_ssec_passthrough_capability(&tgt_client.arn, SsecPassthroughCapability::Unsupported).await; + rinfo.replication_status = ReplicationStatusType::Failed; + rinfo.error = Some(ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED.to_string()); + warn!( + event = EVENT_RESYNC_TARGET_OPERATION_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + object = %object, + arn = %tgt_client.arn, + endpoint = %tgt_client.endpoint, + operation = "ssec_passthrough_audit", + error = ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED, + "Replication target operation failed" + ); + false +} + +/// Post-PUT HEAD-back audit for an SSE-C passthrough replica, over the worker +/// HEAD channel (replication-check exemption plus the `source-proxy-request: +/// false` suppression header, so the target answers locally without a +/// customer key). A HEAD transport failure leaves the capability `Unknown` +/// but still fails this attempt: an unverifiable SSE-C replica must not +/// report COMPLETED. +async fn audit_ssec_passthrough_replica( + tgt_client: &Arc, + bucket: &str, + object: &str, + version_id: Option, + rinfo: &mut ReplicatedTargetInfo, +) -> bool { + // Address the replica the way the PUT named it: a nil source version id + // (versioning-suspended / null-version objects) maps to the "null" + // version, so the audit HEAD does not 4xx-loop on those objects. + let version_id = resolve_read_api_version_id(version_id); + match head_object_for_worker(tgt_client.as_ref(), &tgt_client.bucket, object, version_id).await { + Ok(head) => settle_ssec_passthrough_evidence(&head, tgt_client, bucket, object, rinfo).await, + Err(e) => { + rinfo.replication_status = ReplicationStatusType::Failed; + rinfo.error = Some(format!("SSE-C passthrough audit HEAD failed: {e}")); + warn!( + event = EVENT_RESYNC_TARGET_OPERATION_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + object = %object, + arn = %tgt_client.arn, + operation = "ssec_passthrough_audit_head", + error = %e, + "Replication target operation failed" + ); + mark_replication_target_offline_if_needed(tgt_client, &e).await; + false + } + } +} + +static RESYNC_WORKER_COUNT: usize = 10; + type ResyncCancelKey = (String, String, String); fn configured_resync_max_jobs() -> usize { @@ -632,31 +664,13 @@ impl ReplicationResyncer { } } - #[instrument(skip(cancellation_token, storage))] - pub async fn resync_bucket( - self: Arc, - cancellation_token: CancellationToken, - storage: Arc, - heal: bool, - opts: ResyncOpts, - ) { - // Check cancellation before starting the scan. - // NOTE: the previous design waited here on `worker_rx.resubscribe().recv()` to - // throttle concurrent resyncs, but `resubscribe()` positions the new receiver at - // the current write-head of the broadcast ring buffer, so all pre-sent bootstrap - // signals (written in `ReplicationResyncer::new`) are invisible to it. Every - // spawned task therefore blocked forever, which is why `resync start` reported - // "started" yet objects never moved. Throttling at this level is also incorrect - // for broadcast channels (one send unblocks ALL receivers). The inner - // per-object worker pool (mpsc channels, line ~877) already provides the right - // concurrency limit. - if cancellation_token.is_cancelled() { - return; - } - - // Acquire a cluster-wide leader lock for this (bucket, ARN) pair so that only - // one node runs the resync scan at a time. Without this, every cluster node would - // scan and replicate every object independently, causing N-fold duplicate traffic. + /// Acquire a cluster-wide leader lock for this (bucket, ARN) pair so that only + /// one node runs the resync scan at a time. Without this, every cluster node would + /// scan and replicate every object independently, causing N-fold duplicate traffic. + async fn acquire_resync_leader_lock( + storage: &Arc, + opts: &ResyncOpts, + ) -> Option { let resync_lock_key = ReplicationMetadataStore::resync_lock_key(&opts.bucket, &opts.arn); let resync_ns_lock = match storage .new_ns_lock(ReplicationMetadataStore::rustfs_meta_bucket(), &resync_lock_key) @@ -674,11 +688,11 @@ impl ReplicationResyncer { reason = "leader_lock_create_failed", "Failed to create resync leader lock — skipping resync" ); - return; + return None; } }; - let _resync_leader_guard = match resync_ns_lock.get_write_lock(ReplicationLockTiming::acquire_timeout()).await { - Ok(g) => g, + match resync_ns_lock.get_write_lock(ReplicationLockTiming::acquire_timeout()).await { + Ok(g) => Some(g), Err(_) => { debug!( event = EVENT_RESYNC_STATUS_UPDATE_SKIPPED, @@ -689,14 +703,19 @@ impl ReplicationResyncer { reason = "leader_lock_held_by_another_node", "Another node is already running resync for this bucket/ARN — skipping" ); - return; + None } - }; - - let Some(_resync_admission_permit) = self.acquire_resync_admission(&cancellation_token).await else { - return; - }; + } + } + /// Resolve and validate the replication config plus the single remote target + /// client this resync run replicates to, marking the resync failed (and + /// returning `None`) when any lookup or validation step does not hold. + async fn resolve_resync_target( + &self, + opts: &ResyncOpts, + storage: &Arc, + ) -> Option<(ReplicationConfig, Arc)> { let cfg = match get_replication_config(&opts.bucket).await { Ok(cfg) => cfg, Err(err) => { @@ -712,7 +731,7 @@ impl ReplicationResyncer { ); self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) .await; - return; + return None; } }; @@ -730,7 +749,7 @@ impl ReplicationResyncer { ); self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) .await; - return; + return None; } }; @@ -748,7 +767,7 @@ impl ReplicationResyncer { ); self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) .await; - return; + return None; } let target_arns = if let Some(cfg) = cfg { @@ -773,7 +792,7 @@ impl ReplicationResyncer { ); self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) .await; - return; + return None; } let Some(target_client) = ReplicationTargetStore::remote_target_client(&opts.bucket, &target_arns[0]).await else { @@ -788,9 +807,15 @@ impl ReplicationResyncer { ); self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) .await; - return; + return None; }; + Some((rcfg, target_client)) + } + + /// Persist the `ResyncStarted` status for non-heal runs, logging (without + /// aborting the resync) when the status update fails. + async fn mark_resync_started(&self, heal: bool, opts: &ResyncOpts, storage: &Arc) { if !heal && let Err(e) = self .mark_status(ResyncStatusType::ResyncStarted, opts.clone(), storage.clone()) @@ -807,223 +832,152 @@ impl ReplicationResyncer { "Failed to update resync status" ); } + } - let (tx, mut rx) = tokio::sync::mpsc::channel(100); - let walk_failed = Arc::new(AtomicBool::new(false)); - let walk_failed_task = walk_failed.clone(); - let walk_storage = storage.clone(); - let walk_cancellation = cancellation_token.clone(); - let walk_bucket = opts.bucket.clone(); - let walk_arn = opts.arn.clone(); - let walk_task = tokio::spawn(async move { - if let Err(err) = walk_storage - .walk( - walk_cancellation, - &walk_bucket, - "", - tx, - WalkOptions::default().with_walkdir_timeouts(BACKGROUND_WALKDIR_TIMEOUT), - ) - .await - { - walk_failed_task.store(true, Ordering::Relaxed); - error!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %walk_bucket, - arn = %walk_arn, - reason = "walk_failed", - error = %err, - "Replication resync bucket walk failed" - ); - } - }); + /// Drain and join the resync worker tasks after a fatal dispatch error, + /// logging any observed task failure and persisting the failed status. + async fn finish_resync_failed( + &self, + worker_txs: Vec>, + results_tx: tokio::sync::mpsc::Sender, + futures: Vec>, + join_failure_reason: &str, + opts: &ResyncOpts, + storage: &Arc, + ) { + let worker_failed = finish_resync_workers(worker_txs, results_tx, futures, false).await; + if worker_failed { + error!( + event = EVENT_RESYNC_TASK_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %opts.bucket, + arn = %opts.arn, + reason = join_failure_reason, + "Replication resync worker cleanup observed task failure" + ); + } + self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) + .await; + } - let mut worker_txs = Vec::new(); + /// Abort the resync worker tasks after cancellation and persist the + /// canceled status. + async fn finish_resync_canceled( + &self, + worker_txs: Vec>, + results_tx: tokio::sync::mpsc::Sender, + futures: Vec>, + opts: &ResyncOpts, + storage: &Arc, + ) { + finish_resync_workers(worker_txs, results_tx, futures, true).await; + self.resync_bucket_mark_status(ResyncStatusType::ResyncCanceled, opts.clone(), storage.clone()) + .await; + } + + /// Spawn the collector task that folds per-object resync results into the + /// aggregated resync stats. + fn spawn_resync_results_collector( + resyncer: Arc, + opts: &ResyncOpts, + ) -> (tokio::sync::mpsc::Sender, JoinHandle<()>) { // mpsc, not broadcast: a lagging broadcast receiver returns Err(Lagged) which // would end the collector and silently drop every subsequent worker result. let (results_tx, mut results_rx) = tokio::sync::mpsc::channel::(RESYNC_WORKER_COUNT * 4); let opts_clone = opts.clone(); - let self_clone = self.clone(); - - let mut futures = vec![walk_task]; let results_fut = tokio::spawn(async move { while let Some(st) = results_rx.recv().await { - self_clone.inc_stats(&st, opts_clone.clone()).await; + resyncer.inc_stats(&st, opts_clone.clone()).await; } }); + (results_tx, results_fut) + } + + #[instrument(skip(cancellation_token, storage))] + pub async fn resync_bucket( + self: Arc, + cancellation_token: CancellationToken, + storage: Arc, + heal: bool, + opts: ResyncOpts, + ) { + // Check cancellation before starting the scan. + // NOTE: the previous design waited here on `worker_rx.resubscribe().recv()` to + // throttle concurrent resyncs, but `resubscribe()` positions the new receiver at + // the current write-head of the broadcast ring buffer, so all pre-sent bootstrap + // signals (written in `ReplicationResyncer::new`) are invisible to it. Every + // spawned task therefore blocked forever, which is why `resync start` reported + // "started" yet objects never moved. Throttling at this level is also incorrect + // for broadcast channels (one send unblocks ALL receivers). The inner + // per-object worker pool (mpsc channels, `spawn_resync_object_workers`) already + // provides the right concurrency limit. + if cancellation_token.is_cancelled() { + return; + } + + let Some(_resync_leader_guard) = Self::acquire_resync_leader_lock(&storage, &opts).await else { + return; + }; + + let Some(_resync_admission_permit) = self.acquire_resync_admission(&cancellation_token).await else { + return; + }; + + let Some((rcfg, target_client)) = self.resolve_resync_target(&opts, &storage).await else { + return; + }; + + self.mark_resync_started(heal, &opts, &storage).await; + + let (rx, walk_failed, walk_task) = spawn_resync_walk_task(&storage, &cancellation_token, &opts); + + let mut futures = vec![walk_task]; + + let (results_tx, results_fut) = Self::spawn_resync_results_collector(self.clone(), &opts); + futures.push(results_fut); - for _ in 0..RESYNC_WORKER_COUNT { - let (tx, mut rx) = tokio::sync::mpsc::channel::(100); - worker_txs.push(tx); + let worker_txs = + spawn_resync_object_workers(&cancellation_token, &target_client, &storage, &opts, &results_tx, &mut futures); - let cancel_token = cancellation_token.clone(); - let target_client = target_client.clone(); - let storage = storage.clone(); - let results_tx = results_tx.clone(); - let bucket_name = opts.bucket.clone(); - let target_arn = opts.arn.clone(); + self.drive_resync_dispatch( + &cancellation_token, + rx, + &rcfg, + ResyncRunState { + worker_txs, + results_tx, + futures, + walk_failed, + }, + &opts, + &storage, + ) + .await; + } - let f = tokio::spawn(async move { - while let Some(mut roi) = rx.recv().await { - if cancel_token.is_cancelled() { - return; - } - - if roi.delete_marker || !roi.version_purge_status.is_empty() { - let (version_id, dm_version_id) = if roi.version_purge_status.is_empty() { - (None, roi.version_id) - } else { - (roi.version_id, None) - }; - - let doi = DeletedObjectReplicationInfo { - delete_object: ReplicationDeletedObject { - object_name: roi.name.clone(), - delete_marker_version_id: dm_version_id, - version_id, - replication_state: roi.replication_state.clone(), - delete_marker: roi.delete_marker, - delete_marker_mtime: roi.mod_time, - ..Default::default() - }, - bucket: roi.bucket.clone(), - event_type: REPLICATE_EXISTING_DELETE.to_string(), - op_type: ReplicationType::ExistingObject, - target_arn: target_arn.clone(), - ..Default::default() - }; - replicate_delete(doi, storage.clone()).await; - } else { - roi.op_type = ReplicationType::ExistingObject; - roi.event_type = REPLICATE_EXISTING.to_string(); - replicate_object(roi.clone(), storage.clone()).await; - } - - let mut st = TargetReplicationResyncStatus { - object: roi.name.clone(), - bucket: roi.bucket.clone(), - ..Default::default() - }; - - let reset_id = target_client.reset_id.clone(); - - let head_result = head_object_with_proxy_stats( - &bucket_name, - target_client.as_ref(), - &target_client.bucket, - &roi.name, - roi.version_id.map(|v| v.to_string()), - ) - .await; - let (size, err) = match head_result { - Ok(_) => { - st.replicated_count += 1; - st.replicated_size += roi.size; - (roi.size, None) - } - Err(err) if roi.delete_marker => { - // Verifying a replicated delete marker: only a - // definitive 404/NoSuchKey or 405/MethodNotAllowed - // confirms the marker propagated. Any other - // (retryable/ambiguous) HEAD error leaves the outcome - // unverified, so it must count as failed — not as a - // blanket success (backlog#862 / #799 B13). - let retryable = { - let (is_not_found, code) = err - .as_service_error() - .map(|se| (se.is_not_found(), se.code())) - .unwrap_or((false, None)); - is_retryable_delete_replication_head_error(is_not_found, code) - }; - if retryable { - st.failed_count += 1; - (0, Some(err)) - } else { - st.replicated_count += 1; - (0, None) - } - } - Err(err) if is_version_id_format_mismatch(&err) => { - // AWS-style target rejects the RustFS UUID versionId - // (400). Re-verify without the versionId before - // concluding the object failed to replicate, instead - // of counting a well-replicated object as failed. - match head_object_fallback(&bucket_name, target_client.as_ref(), &roi.name).await { - Ok(Some(_)) => { - st.replicated_count += 1; - st.replicated_size += roi.size; - (roi.size, None) - } - Ok(None) => { - st.failed_count += 1; - (0, Some(err)) - } - Err(e2) => { - st.failed_count += 1; - (0, Some(e2)) - } - } - } - Err(err) => { - st.failed_count += 1; - (0, Some(err)) - } - }; - - if err.is_some() { - debug!( - event = EVENT_RESYNC_OBJECT_PROCESSED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - reset_id = %reset_id, - bucket = %bucket_name, - object = %roi.name, - version_id = %roi.version_id.unwrap_or_default(), - size, - error = ?err, - "Processed resync object with verification error" - ); - } else { - trace!( - event = EVENT_RESYNC_OBJECT_PROCESSED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - reset_id = %reset_id, - bucket = %bucket_name, - object = %roi.name, - version_id = %roi.version_id.unwrap_or_default(), - size, - "Processed resync object" - ); - } - st.error = err.as_ref().and_then(resync_target_error_detail); - - if cancel_token.is_cancelled() { - return; - } - - if let Err(err) = results_tx.send(st).await { - error!( - event = EVENT_RESYNC_RUNTIME_CHANNEL_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket_name, - reason = "status_channel_send_failed", - error = %err, - "Failed to send resync status" - ); - } - } - }); - - futures.push(f); - } + /// Pump walked objects through classification into the hashed worker + /// queues, finalizing the resync status on dispatch error, cancellation, + /// or completion of the walk. + async fn drive_resync_dispatch( + &self, + cancellation_token: &CancellationToken, + mut rx: tokio::sync::mpsc::Receiver>, + rcfg: &ReplicationConfig, + state: ResyncRunState, + opts: &ResyncOpts, + storage: &Arc, + ) { + let ResyncRunState { + worker_txs, + results_tx, + futures, + walk_failed, + } = state; while let Some(res) = rx.recv().await { if let Some(err) = res.err { @@ -1039,27 +993,21 @@ impl ReplicationResyncer { ); cancellation_token.cancel(); drop(rx); - let worker_failed = finish_resync_workers(worker_txs, results_tx, futures, false).await; - if worker_failed { - error!( - event = EVENT_RESYNC_TASK_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %opts.bucket, - arn = %opts.arn, - reason = "worker_join_failed_after_object_info_error", - "Replication resync worker cleanup observed task failure" - ); - } - self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) - .await; + self.finish_resync_failed( + worker_txs, + results_tx, + futures, + "worker_join_failed_after_object_info_error", + opts, + storage, + ) + .await; return; } if cancellation_token.is_cancelled() { drop(rx); - finish_resync_workers(worker_txs, results_tx, futures, true).await; - self.resync_bucket_mark_status(ResyncStatusType::ResyncCanceled, opts.clone(), storage.clone()) + self.finish_resync_canceled(worker_txs, results_tx, futures, opts, storage) .await; return; } @@ -1068,7 +1016,7 @@ impl ReplicationResyncer { continue; }; - let roi = match get_heal_replicate_object_info(&object, &rcfg).await { + let roi = match get_heal_replicate_object_info(&object, rcfg).await { Ok(roi) => roi, Err(err) => { error!( @@ -1083,20 +1031,15 @@ impl ReplicationResyncer { ); cancellation_token.cancel(); drop(rx); - let worker_failed = finish_resync_workers(worker_txs, results_tx, futures, false).await; - if worker_failed { - error!( - event = EVENT_RESYNC_TASK_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %opts.bucket, - arn = %opts.arn, - reason = "worker_join_failed_after_classification_error", - "Replication resync worker cleanup observed task failure" - ); - } - self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) - .await; + self.finish_resync_failed( + worker_txs, + results_tx, + futures, + "worker_join_failed_after_classification_error", + opts, + storage, + ) + .await; return; } }; @@ -1106,8 +1049,7 @@ impl ReplicationResyncer { if cancellation_token.is_cancelled() { drop(rx); - finish_resync_workers(worker_txs, results_tx, futures, true).await; - self.resync_bucket_mark_status(ResyncStatusType::ResyncCanceled, opts.clone(), storage.clone()) + self.finish_resync_canceled(worker_txs, results_tx, futures, opts, storage) .await; return; } @@ -1127,26 +1069,21 @@ impl ReplicationResyncer { ); cancellation_token.cancel(); drop(rx); - let worker_failed = finish_resync_workers(worker_txs, results_tx, futures, false).await; - if worker_failed { - error!( - event = EVENT_RESYNC_TASK_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %opts.bucket, - arn = %opts.arn, - reason = "worker_join_failed_after_queue_send_error", - "Replication resync worker cleanup observed task failure" - ); - } - self.resync_bucket_mark_status(ResyncStatusType::ResyncFailed, opts.clone(), storage.clone()) - .await; + self.finish_resync_failed( + worker_txs, + results_tx, + futures, + "worker_join_failed_after_queue_send_error", + opts, + storage, + ) + .await; return; } } let worker_failed = finish_resync_workers(worker_txs, results_tx, futures, false).await; - let target_failed = self.target_has_resync_failures(&opts).await; + let target_failed = self.target_has_resync_failures(opts).await; let status = if walk_failed.load(Ordering::Relaxed) || worker_failed || target_failed { ResyncStatusType::ResyncFailed } else { @@ -1157,6 +1094,247 @@ impl ReplicationResyncer { } } +/// Worker-pool channel and task state for one resync run, handed from setup to +/// the dispatch loop. +struct ResyncRunState { + worker_txs: Vec>, + results_tx: tokio::sync::mpsc::Sender, + futures: Vec>, + walk_failed: Arc, +} + +/// Spawn the bucket walk task that feeds object listings into the resync +/// dispatch loop, surfacing walk failures through the returned flag. +fn spawn_resync_walk_task( + storage: &Arc, + cancellation_token: &CancellationToken, + opts: &ResyncOpts, +) -> ( + tokio::sync::mpsc::Receiver>, + Arc, + JoinHandle<()>, +) { + let (tx, rx) = tokio::sync::mpsc::channel(100); + let walk_failed = Arc::new(AtomicBool::new(false)); + let walk_failed_task = walk_failed.clone(); + let walk_storage = storage.clone(); + let walk_cancellation = cancellation_token.clone(); + let walk_bucket = opts.bucket.clone(); + let walk_arn = opts.arn.clone(); + let walk_task = tokio::spawn(async move { + if let Err(err) = walk_storage + .walk( + walk_cancellation, + &walk_bucket, + "", + tx, + WalkOptions::default().with_walkdir_timeouts(BACKGROUND_WALKDIR_TIMEOUT), + ) + .await + { + walk_failed_task.store(true, Ordering::Relaxed); + error!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %walk_bucket, + arn = %walk_arn, + reason = "walk_failed", + error = %err, + "Replication resync bucket walk failed" + ); + } + }); + (rx, walk_failed, walk_task) +} + +/// Classify the target HEAD verification result for one resynced object, +/// updating the per-object status counters and returning the accounted size +/// together with any verification error. +async fn verify_resync_head_result( + head_result: std::result::Result>, + roi: &ReplicateObjectInfo, + st: &mut TargetReplicationResyncStatus, + target_client: &Arc, +) -> (i64, Option>) { + match head_result { + Ok(_) => { + st.replicated_count += 1; + st.replicated_size += roi.size; + (roi.size, None) + } + Err(err) if roi.delete_marker => { + // Verifying a replicated delete marker: only a + // definitive 404/NoSuchKey or 405/MethodNotAllowed + // confirms the marker propagated. Any other + // (retryable/ambiguous) HEAD error leaves the outcome + // unverified, so it must count as failed — not as a + // blanket success (backlog#862 / #799 B13). + let retryable = { + let (is_not_found, code) = err + .as_service_error() + .map(|se| (se.is_not_found(), se.code())) + .unwrap_or((false, None)); + is_retryable_delete_replication_head_error(is_not_found, code) + }; + if retryable { + st.failed_count += 1; + (0, Some(err)) + } else { + st.replicated_count += 1; + (0, None) + } + } + Err(err) if is_version_id_format_mismatch(&err) => { + // AWS-style target rejects the RustFS UUID versionId + // (400). Re-verify without the versionId before + // concluding the object failed to replicate, instead + // of counting a well-replicated object as failed. + match head_object_fallback(target_client.as_ref(), &roi.name).await { + Ok(Some(_)) => { + st.replicated_count += 1; + st.replicated_size += roi.size; + (roi.size, None) + } + Ok(None) => { + st.failed_count += 1; + (0, Some(err)) + } + Err(e2) => { + st.failed_count += 1; + (0, Some(e2)) + } + } + } + Err(err) => { + st.failed_count += 1; + (0, Some(err)) + } + } +} + +/// Replicate one existing object (or delete marker / version purge) to the +/// resync target, verify the outcome via a target HEAD, and produce the +/// per-object resync status update. +async fn resync_worker_process_object( + mut roi: ReplicateObjectInfo, + storage: &Arc, + target_client: &Arc, + bucket_name: &str, + target_arn: &str, +) -> TargetReplicationResyncStatus { + if roi.delete_marker || !roi.version_purge_status.is_empty() { + let doi = resync_existing_delete_replication_info(&roi, target_arn); + replicate_delete(doi, storage.clone()).await; + } else { + roi.op_type = ReplicationType::ExistingObject; + roi.event_type = REPLICATE_EXISTING.to_string(); + replicate_object(roi.clone(), storage.clone()).await; + } + + let mut st = TargetReplicationResyncStatus { + object: roi.name.clone(), + bucket: roi.bucket.clone(), + ..Default::default() + }; + + let reset_id = target_client.reset_id.clone(); + + let head_result = head_object_for_worker( + target_client.as_ref(), + &target_client.bucket, + &roi.name, + roi.version_id.map(|v| v.to_string()), + ) + .await; + let (size, err) = verify_resync_head_result(head_result, &roi, &mut st, target_client).await; + + if err.is_some() { + debug!( + event = EVENT_RESYNC_OBJECT_PROCESSED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + reset_id = %reset_id, + bucket = %bucket_name, + object = %roi.name, + version_id = %roi.version_id.unwrap_or_default(), + size, + error = ?err, + "Processed resync object with verification error" + ); + } else { + trace!( + event = EVENT_RESYNC_OBJECT_PROCESSED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + reset_id = %reset_id, + bucket = %bucket_name, + object = %roi.name, + version_id = %roi.version_id.unwrap_or_default(), + size, + "Processed resync object" + ); + } + st.error = err.as_ref().and_then(resync_target_error_detail); + + st +} + +/// Spawn the per-object resync worker pool, wiring every worker to the shared +/// results channel and registering its task handle for cleanup. +fn spawn_resync_object_workers( + cancellation_token: &CancellationToken, + target_client: &Arc, + storage: &Arc, + opts: &ResyncOpts, + results_tx: &tokio::sync::mpsc::Sender, + futures: &mut Vec>, +) -> Vec> { + let mut worker_txs = Vec::new(); + + for _ in 0..RESYNC_WORKER_COUNT { + let (tx, mut rx) = tokio::sync::mpsc::channel::(100); + worker_txs.push(tx); + + let cancel_token = cancellation_token.clone(); + let target_client = target_client.clone(); + let storage = storage.clone(); + let results_tx = results_tx.clone(); + let bucket_name = opts.bucket.clone(); + let target_arn = opts.arn.clone(); + + let f = tokio::spawn(async move { + while let Some(roi) = rx.recv().await { + if cancel_token.is_cancelled() { + return; + } + + let st = resync_worker_process_object(roi, &storage, &target_client, &bucket_name, &target_arn).await; + + if cancel_token.is_cancelled() { + return; + } + + if let Err(err) = results_tx.send(st).await { + error!( + event = EVENT_RESYNC_RUNTIME_CHANNEL_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket_name, + reason = "status_channel_send_failed", + error = %err, + "Failed to send resync status" + ); + } + } + }); + + futures.push(f); + } + + worker_txs +} + pub async fn get_heal_replicate_object_info(oi: &ObjectInfo, rcfg: &ReplicationConfig) -> Result { let mut oi = oi.clone(); let mut user_defined = (*oi.user_defined).clone(); @@ -1686,29 +1864,6 @@ pub(crate) async fn replicate_delete_with_outcome( ) } -/// Whether a delete replication fully succeeded — the MRF replay acknowledges -/// (drops) an entry exactly when this returns true. -/// -/// The delayed purge is deliberately NOT an input: holding the outcome hostage -/// to it (`&& !requires_delayed_purge`) forced `false` for every delete-marker -/// entry and retained them all in the durable MRF journal forever. Purge -/// failures persist their own purge-intent entry instead -/// (`watch_and_purge_source_delete_marker`), and replays of those entries -/// report purge success through `purge_stale_delete_marker_targets`. -fn replicate_delete_outcome( - expected_targets: usize, - replicated_targets: usize, - state_persisted: bool, - source_state_verified: bool, - replication_status: &ReplicationStatusType, -) -> bool { - expected_targets > 0 - && replicated_targets == expected_targets - && state_persisted - && source_state_verified - && *replication_status == ReplicationStatusType::Completed -} - async fn source_delete_marker_missing( storage: &S, bucket: &str, @@ -1733,29 +1888,6 @@ async fn source_delete_marker_missing( } } -/// Which version a delete-marker purge should address on one target. -/// -/// `None` means do not purge at all: the recorded mapping disagreed across the -/// dual internal prefixes, and guessing an id could destroy a live version on -/// the target. `Some(id)` is the exact version the target reported when it -/// accepted the marker; falling back to a source-derived id is only correct -/// when the target mirrors source version ids, which a generic S3 target does -/// not. -fn delete_marker_purge_version_id( - state: Option<&ReplicationState>, - arn: &str, - delete_marker_version_id: Uuid, -) -> Option> { - if state.is_some_and(|state| state.target_delete_marker_version_ids_corrupt) { - return None; - } - let recorded = state.and_then(|state| state.target_delete_marker_version_ids.get(arn).cloned()); - Some(match recorded { - Some(version_id) => Some(version_id), - None => target_delete_version_id(delete_marker_version_id, true), - }) -} - /// One purge pass over the eligible targets. Returns the ARNs that must be /// retried: the remote DELETE failed, or the target client was unavailable /// (e.g. a runtime cache miss). Inconsistent recorded version mappings are a @@ -1923,20 +2055,6 @@ async fn watch_and_purge_source_delete_marker( } } -/// Shape an exhausted purge intent as a marker-creation delete entry. Replay -/// reconstructs it with `delete_marker: true`, finds the source marker gone, -/// and funnels into the stale-marker branch of `replicate_delete_with_outcome` -/// — which re-runs the purge without touching source state and reports purge -/// success as the replay outcome. -fn delete_marker_purge_mrf_entry(dobj: &DeletedObjectReplicationInfo, failed_arns: Vec) -> MrfReplicateEntry { - let mut entry = dobj.to_mrf_entry(); - entry.delete_marker = true; - entry.version_id = None; - entry.retry_count = 0; - entry.target_arns = failed_arns; - entry -} - async fn enqueue_delete_marker_purge_mrf(dobj: &DeletedObjectReplicationInfo, failed_arns: Vec) { let arns = failed_arns.join(","); let miss_reason = match runtime_sources::replication_pool() { @@ -2296,14 +2414,6 @@ async fn replicate_force_delete_to_targets(dobj: &Deleted all_succeeded } -fn target_delete_version_id(version_id: Uuid, version_purge: bool) -> Option { - if version_id.is_nil() { - version_purge.then(|| NULL_VERSION_ID.to_string()) - } else { - Some(version_id.to_string()) - } -} - async fn replicate_delete_to_target(dobj: &DeletedObjectReplicationInfo, tgt_client: Arc) -> ReplicatedTargetInfo { let version_id = if let Some(version_id) = &dobj.delete_object.delete_marker_version_id { version_id.to_owned() @@ -2346,8 +2456,7 @@ async fn replicate_delete_to_target(dobj: &DeletedObjectReplicationInfo, tgt_cli let version_id = target_delete_version_id(version_id, is_version_purge); if dobj.delete_object.delete_marker && dobj.delete_object.delete_marker_version_id.is_some() { - match head_object_with_proxy_stats( - &dobj.bucket, + match head_object_for_worker( tgt_client.as_ref(), &tgt_client.bucket, &dobj.delete_object.object_name, @@ -2713,6 +2822,21 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { return rinfo; } + // N2 fail-closed: never PUT SSE-C ciphertext at a target known to drop + // the passthrough transport headers, and never trust a convergence HEAD + // against such a target — a previous broken replica matches by ETag. + let Some(ssec_audit_required) = resolve_ssec_passthrough_gate(self.ssec, &tgt_client, &bucket, &object, &mut rinfo).await + else { + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.clone(), + object: self.to_object_info(), + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + return rinfo; + }; + let versioned = ReplicationVersioningStore::prefix_enabled(&bucket, &object).await; let version_suspended = ReplicationVersioningStore::prefix_suspended(&bucket, &object).await; @@ -2810,18 +2934,20 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { } let mut replication_action = replication_action; - match head_object_with_proxy_stats( - &bucket, - tgt_client.as_ref(), - &tgt_client.bucket, - &object, - self.version_id.map(|v| v.to_string()), - ) - .await + match head_object_for_worker(tgt_client.as_ref(), &tgt_client.bucket, &object, self.version_id.map(|v| v.to_string())) + .await { Ok(oi) => { replication_action = replication_action_for_target_head(&object_info, &oi, self.op_type); if replication_action == ReplicationAction::None { + // An SSE-C replica only counts as converged when the same + // HEAD proves its decryption material survived; a broken + // ciphertext copy from an earlier attempt matches by ETag. + if ssec_audit_required + && !settle_ssec_passthrough_evidence(&oi, &tgt_client, &bucket, &object, &mut rinfo).await + { + return rinfo; + } rinfo.replication_status = ReplicationStatusType::Completed; rinfo.replication_resynced = true; rinfo.replication_action = ReplicationAction::None; @@ -2834,8 +2960,13 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { // Object not on target yet → fall through to PUT. } else if is_version_id_format_mismatch(&e) { // Version-ID format mismatch: retry without versionId and compare ETags. - match head_object_fallback(&bucket, &tgt_client, &object).await { + match head_object_fallback(&tgt_client, &object).await { Ok(Some(oi)) if replication_etags_match(object_info.etag.as_deref(), oi.e_tag.as_deref()) => { + if ssec_audit_required + && !settle_ssec_passthrough_evidence(&oi, &tgt_client, &bucket, &object, &mut rinfo).await + { + return rinfo; + } rinfo.replication_status = ReplicationStatusType::Completed; rinfo.replication_resynced = true; rinfo.replication_action = ReplicationAction::None; @@ -2910,7 +3041,6 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { } }; - let has_tagging_replication = !put_opts.user_tags.is_empty(); if let Some(err) = if is_multipart { drop(gr); let result = replicate_object_with_multipart(MultipartReplicationContext { @@ -2925,10 +3055,6 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { put_opts, }) .await; - record_proxy_request(&bucket, "PutObject", result.is_err()).await; - if has_tagging_replication { - record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await; - } result.err() } else { gr.stream = wrap_with_bandwidth_monitor(gr.stream, &put_opts, &bucket, &rinfo.arn); @@ -2944,10 +3070,6 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { ) }) .map_err(|e| std::io::Error::other(e.to_string())); - record_proxy_request(&bucket, "PutObject", result.is_err()).await; - if has_tagging_replication { - record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await; - } result.err() } { rinfo.replication_status = ReplicationStatusType::Failed; @@ -2969,6 +3091,15 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { return rinfo; } + // First SSE-C passthrough PUT against this target: verify the replica + // kept its decryption material before reporting COMPLETED. + if ssec_audit_required + && !audit_ssec_passthrough_replica(&tgt_client, &bucket, &object, self.version_id.map(|v| v.to_string()), &mut rinfo) + .await + { + return rinfo; + } + rinfo.replication_status = ReplicationStatusType::Completed; rinfo @@ -2984,30 +3115,17 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { let bucket = self.bucket.clone(); let object = self.name.clone(); - let mut replication_action = ReplicationAction::Metadata; - let mut rinfo = ReplicatedTargetInfo { - arn: tgt_client.arn.clone(), - size: self.actual_size, - replication_action, - op_type: self.op_type, - replication_status: ReplicationStatusType::Failed, - prev_replication_status: self.target_replication_status(&tgt_client.arn), - endpoint: tgt_client.endpoint.clone(), - secure: tgt_client.secure, - ..Default::default() - }; + let mut rinfo = replicate_all_target_info(self, &tgt_client); if ReplicationTargetStore::target_is_offline(&tgt_client).await { - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - target = %tgt_client.to_url(), - reason = "target_offline", - "Skipped replication because target is offline" - ); + note_replicate_all_target_offline(self, &bucket, &tgt_client); + return rinfo; + } + + // N2 fail-closed: see the gate in `replicate_object` — the same policy + // applies to the metadata/existing-object transport. + let Some(ssec_audit_required) = resolve_ssec_passthrough_gate(self.ssec, &tgt_client, &bucket, &object, &mut rinfo).await + else { send_local_event(EventArgs { event_name: EventName::ObjectReplicationNotTracked.to_string(), bucket_name: bucket.clone(), @@ -3015,49 +3133,22 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { user_agent: "Internal: [Replication]".to_string(), ..Default::default() }); + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); return rinfo; - } + }; let versioned = ReplicationVersioningStore::prefix_enabled(&bucket, &object).await; let version_suspended = ReplicationVersioningStore::prefix_suspended(&bucket, &object).await; - let obj_opts = ObjectOptions { - version_id: self.version_id.map(|v| v.to_string()), - version_suspended, - versioned, - replication_request: true, - // SSE-C passthrough reads the stored ciphertext verbatim; the - // decrypting reader cannot serve it (no customer key server-side). - raw_data_movement_read: self.ssec, - ..Default::default() - }; + let obj_opts = replicate_all_read_options(self, versioned, version_suspended); - let mut gr = match storage + let gr = match storage .get_object_reader(&bucket, &object, None, HeaderMap::new(), &obj_opts) .await { Ok(gr) => gr, Err(e) => { - if !(is_err_object_not_found(&e) || is_err_version_not_found(&e)) { - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - error = %e, - reason = "object_reader_unavailable", - "Skipped replication because object reader is unavailable" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: self.to_object_info(), - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); - } - + note_replicate_all_reader_unavailable(self, &bucket, &tgt_client, &e); return rinfo; } }; @@ -3069,23 +3160,7 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { let size = match object_info.get_actual_size() { Ok(size) => size, Err(e) => { - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - error = %e, - reason = "actual_size_unavailable", - "Skipped replication because actual object size is unavailable" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info, - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); + note_replicate_all_size_unavailable(&bucket, &tgt_client, object_info, &e); return rinfo; } }; @@ -3094,174 +3169,27 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { let transfer_size = if self.ssec { object_info.size } else { size }; if tgt_client.bucket.is_empty() { - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - reason = "target_bucket_empty", - "Skipped replication because target bucket is empty" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info, - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); + note_replicate_all_target_bucket_empty(&bucket, &tgt_client, object_info); return rinfo; } - let mut sopts = StatObjectOptions { - version_id: object_info.version_id.map(|v| v.to_string()).unwrap_or_default(), - internal: AdvancedGetOptions { - replication_proxy_request: "false".to_string(), - ..Default::default() + let _sopts = replicate_all_stat_options(&object_info, &bucket, &tgt_client); + + let Some((replication_action, object_info)) = resolve_replicate_all_action( + ReplicateAllActionContext { + roi: self, + tgt_client: &tgt_client, + bucket: &bucket, + object: &object, + start_time, + ssec_audit_required, }, - ..Default::default() - }; - - if let Err(err) = sopts.set(AMZ_TAGGING_DIRECTIVE, "ACCESS") { - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - error = %err, - reason = "tagging_directive_header_invalid", - "Skipped replication tagging directive header detail" - ); - } - - match head_object_with_proxy_stats( - &bucket, - tgt_client.as_ref(), - &tgt_client.bucket, - &object, - self.version_id.map(|v| v.to_string()), + object_info, + &mut rinfo, ) .await - { - Ok(oi) => { - replication_action = replication_action_for_target_head(&object_info, &oi, self.op_type); - rinfo.replication_status = ReplicationStatusType::Completed; - if replication_action == ReplicationAction::None { - if self.op_type == ReplicationType::ExistingObject - && replication_target_head_is_newer_null_version(&object_info, &oi) - { - warn!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - object = %object, - arn = %tgt_client.arn, - endpoint = %tgt_client.to_url(), - reason = "target_newer_than_source_null_version", - "Skipping replication because newer target version exists" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info.clone(), - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); - } - - if object_info.target_replication_status(&tgt_client.arn) == ReplicationStatusType::Pending - || object_info.target_replication_status(&tgt_client.arn) == ReplicationStatusType::Failed - || self.op_type == ReplicationType::ExistingObject - { - rinfo.replication_action = replication_action; - rinfo.replication_status = ReplicationStatusType::Completed; - } - - if rinfo.replication_status == ReplicationStatusType::Completed - && self.op_type == ReplicationType::ExistingObject - && !tgt_client.reset_id.is_empty() - { - rinfo.resync_timestamp = format!( - "{};{}", - OffsetDateTime::now_utc() - .format(&Rfc3339) - .unwrap_or_else(|_| "invalid-time".to_string()), - tgt_client.reset_id - ); - rinfo.replication_resynced = true; - } - - rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); - - return rinfo; - } - } - Err(e) => { - if is_version_id_format_mismatch(&e) { - // Version-ID format mismatch: retry without versionId and compare ETags. - match head_object_fallback(&bucket, &tgt_client, &object).await { - Ok(Some(oi)) => { - replication_action = if replication_etags_match(object_info.etag.as_deref(), oi.e_tag.as_deref()) { - ReplicationAction::None - } else { - ReplicationAction::All - }; - } - Ok(None) => { - replication_action = ReplicationAction::All; - } - Err(e2) => { - rinfo.error = Some(e2.to_string()); - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - error = %e2, - reason = "head_object_fallback_failed", - "Failed replication head-object fallback" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info, - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); - rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); - return rinfo; - } - } - } else if e.as_service_error().is_some_and(|se| se.is_not_found()) { - replication_action = ReplicationAction::All; - } else { - rinfo.error = Some(e.to_string()); - debug!( - event = EVENT_RESYNC_RUNTIME_SKIPPED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - error = %e, - reason = "head_object_failed", - "Skipped replication because head-object failed" - ); - - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info, - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); - - rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); - return rinfo; - } - } + else { + return rinfo; }; rinfo.replication_status = ReplicationStatusType::Completed; @@ -3276,14 +3204,7 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { // AWS-style targets permanently FAILED and never converging // (backlog#860 / #799 B11). if self.op_type == ReplicationType::ExistingObject && !tgt_client.reset_id.is_empty() { - rinfo.resync_timestamp = format!( - "{};{}", - OffsetDateTime::now_utc() - .format(&Rfc3339) - .unwrap_or_else(|_| "invalid-time".to_string()), - tgt_client.reset_id - ); - rinfo.replication_resynced = true; + apply_replication_resync_timestamp(&mut rinfo, &tgt_client.reset_id); } rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); return rinfo; @@ -3295,91 +3216,39 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { let (put_opts, is_multipart) = match replication_put_object_options(&tgt_client.storage_class, &object_info) { Ok((put_opts, is_mp)) => (put_opts, is_mp), Err(e) => { - // Unsupported source metadata (e.g. managed SSE) is a fail-closed - // condition: report FAILED so the composite status and the - // OperationFailedReplication event reflect that nothing reached - // the target, instead of leaking the optimistic Completed above. - rinfo.replication_status = ReplicationStatusType::Failed; - rinfo.error = Some(e.to_string()); - warn!( - event = EVENT_RESYNC_TARGET_OPERATION_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - operation = "build_put_options", - error = %e, - "Replication target operation failed" - ); - send_local_event(EventArgs { - event_name: EventName::ObjectReplicationNotTracked.to_string(), - bucket_name: bucket.clone(), - object: object_info, - user_agent: "Internal: [Replication]".to_string(), - ..Default::default() - }); - - rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + fail_replicate_all_put_options(&mut rinfo, &tgt_client, &bucket, object_info, &e, start_time); return rinfo; } }; - let has_tagging_replication = !put_opts.user_tags.is_empty(); - if let Some(err) = if is_multipart { - drop(gr); - let result = replicate_object_with_multipart(MultipartReplicationContext { - storage: storage.clone(), - cli: tgt_client.clone(), - src_bucket: &bucket, - dst_bucket: &tgt_client.bucket, + if let Some(err) = replicate_all_payload_to_target( + ReplicateAllPayloadContext { + storage: &storage, + tgt_client: &tgt_client, + bucket: &bucket, object: &object, object_info: &object_info, obj_opts: &obj_opts, arn: &rinfo.arn, + transfer_size, + is_multipart, put_opts, - }) - .await; - record_proxy_request(&bucket, "PutObject", result.is_err()).await; - if has_tagging_replication { - record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await; - } - result.err() - } else { - gr.stream = wrap_with_bandwidth_monitor(gr.stream, &put_opts, &bucket, &rinfo.arn); - let byte_stream = async_read_to_bytestream(gr.stream); - let result = tgt_client - .put_object(&tgt_client.bucket, &object, transfer_size, byte_stream, &put_opts) - .await - .map(|assigned_version_id| { - audit_target_version_identity( - &tgt_client, - &put_opts.internal.source_version_id, - assigned_version_id.as_deref(), - ) - }) - .map_err(|e| std::io::Error::other(e.to_string())); - record_proxy_request(&bucket, "PutObject", result.is_err()).await; - if has_tagging_replication { - record_proxy_request(&bucket, "PutObjectTagging", result.is_err()).await; - } - result.err() - } { - rinfo.replication_status = ReplicationStatusType::Failed; - rinfo.error = Some(err.to_string()); - warn!( - event = EVENT_RESYNC_TARGET_OPERATION_FAILED, - component = LOG_COMPONENT_ECSTORE, - subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, - bucket = %bucket, - arn = %tgt_client.arn, - object = %object, - operation = "put_object", - error = ?err, - "Replication target operation failed" - ); - rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + }, + gr, + ) + .await + { + fail_replicate_all_put_object(&mut rinfo, &tgt_client, &bucket, &object, &err, start_time).await; + return rinfo; + } - mark_replication_target_offline_if_needed(&tgt_client, &err).await; + // First SSE-C passthrough PUT against this target: verify the replica + // kept its decryption material before reporting COMPLETED. + if ssec_audit_required + && !audit_ssec_passthrough_replica(&tgt_client, &bucket, &object, self.version_id.map(|v| v.to_string()), &mut rinfo) + .await + { + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); return rinfo; } @@ -3406,6 +3275,447 @@ impl ReplicateObjectInfoExt for ReplicateObjectInfo { } } +/// Build the initial replication outcome DTO for `replicate_all`, seeded with +/// the metadata-only action and a failed status until the target confirms +/// otherwise. +fn replicate_all_target_info(roi: &ReplicateObjectInfo, tgt_client: &TargetClient) -> ReplicatedTargetInfo { + ReplicatedTargetInfo { + arn: tgt_client.arn.clone(), + size: roi.actual_size, + replication_action: ReplicationAction::Metadata, + op_type: roi.op_type, + replication_status: ReplicationStatusType::Failed, + prev_replication_status: roi.target_replication_status(&tgt_client.arn), + endpoint: tgt_client.endpoint.clone(), + secure: tgt_client.secure, + ..Default::default() + } +} + +/// Log and notify that replication was skipped because the target is offline. +fn note_replicate_all_target_offline(roi: &ReplicateObjectInfo, bucket: &str, tgt_client: &TargetClient) { + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + target = %tgt_client.to_url(), + reason = "target_offline", + "Skipped replication because target is offline" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: roi.to_object_info(), + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); +} + +/// Build the source-side read options for `replicate_all`. +fn replicate_all_read_options(roi: &ReplicateObjectInfo, versioned: bool, version_suspended: bool) -> ObjectOptions { + ObjectOptions { + version_id: roi.version_id.map(|v| v.to_string()), + version_suspended, + versioned, + replication_request: true, + // SSE-C passthrough reads the stored ciphertext verbatim; the + // decrypting reader cannot serve it (no customer key server-side). + raw_data_movement_read: roi.ssec, + ..Default::default() + } +} + +/// Log and notify that replication was skipped because the source object +/// reader is unavailable; missing objects/versions stay silent. +fn note_replicate_all_reader_unavailable(roi: &ReplicateObjectInfo, bucket: &str, tgt_client: &TargetClient, e: &Error) { + if !(is_err_object_not_found(e) || is_err_version_not_found(e)) { + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + error = %e, + reason = "object_reader_unavailable", + "Skipped replication because object reader is unavailable" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: roi.to_object_info(), + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + } +} + +/// Log and notify that replication was skipped because the actual object size +/// is unavailable. +fn note_replicate_all_size_unavailable(bucket: &str, tgt_client: &TargetClient, object_info: ObjectInfo, e: &std::io::Error) { + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + error = %e, + reason = "actual_size_unavailable", + "Skipped replication because actual object size is unavailable" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info, + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); +} + +/// Log and notify that replication was skipped because the target bucket is +/// empty. +fn note_replicate_all_target_bucket_empty(bucket: &str, tgt_client: &TargetClient, object_info: ObjectInfo) { + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + reason = "target_bucket_empty", + "Skipped replication because target bucket is empty" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info, + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); +} + +/// Build the stat options for the target metadata comparison, logging (without +/// failing) when the tagging directive header cannot be set. +fn replicate_all_stat_options(object_info: &ObjectInfo, bucket: &str, tgt_client: &TargetClient) -> StatObjectOptions { + let mut sopts = StatObjectOptions { + version_id: object_info.version_id.map(|v| v.to_string()).unwrap_or_default(), + internal: AdvancedGetOptions { + replication_proxy_request: "false".to_string(), + ..Default::default() + }, + ..Default::default() + }; + + if let Err(err) = sopts.set(AMZ_TAGGING_DIRECTIVE, "ACCESS") { + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + error = %err, + reason = "tagging_directive_header_invalid", + "Skipped replication tagging directive header detail" + ); + } + + sopts +} + +/// Record a failed payload transfer: mark the outcome FAILED, log the target +/// operation failure, and take the target offline when the error is a network +/// failure. +async fn fail_replicate_all_put_object( + rinfo: &mut ReplicatedTargetInfo, + tgt_client: &Arc, + bucket: &str, + object: &str, + err: &std::io::Error, + start_time: OffsetDateTime, +) { + rinfo.replication_status = ReplicationStatusType::Failed; + rinfo.error = Some(err.to_string()); + warn!( + event = EVENT_RESYNC_TARGET_OPERATION_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + object = %object, + operation = "put_object", + error = ?err, + "Replication target operation failed" + ); + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + + mark_replication_target_offline_if_needed(tgt_client, err).await; +} + +/// Stamp the replication outcome as resynced against the target's current +/// reset id. +fn apply_replication_resync_timestamp(rinfo: &mut ReplicatedTargetInfo, reset_id: &str) { + rinfo.resync_timestamp = format!( + "{};{}", + OffsetDateTime::now_utc() + .format(&Rfc3339) + .unwrap_or_else(|_| "invalid-time".to_string()), + reset_id + ); + rinfo.replication_resynced = true; +} + +/// Borrowed inputs for [`resolve_replicate_all_action`]. +struct ReplicateAllActionContext<'a> { + roi: &'a ReplicateObjectInfo, + tgt_client: &'a Arc, + bucket: &'a str, + object: &'a str, + start_time: OffsetDateTime, + /// N2: the target's SSE-C passthrough capability is still `Unknown`, so a + /// converged-looking replica must additionally prove its SSE-C material + /// survived before the comparison may settle COMPLETED. + ssec_audit_required: bool, +} + +/// Compare the source object against the target via HEAD and decide which +/// replication action is still required. Returns `None` after fully settling +/// `rinfo` when replication must stop here — either because the target already +/// matches or because the comparison failed. +async fn resolve_replicate_all_action( + ctx: ReplicateAllActionContext<'_>, + object_info: ObjectInfo, + rinfo: &mut ReplicatedTargetInfo, +) -> Option<(ReplicationAction, ObjectInfo)> { + let ReplicateAllActionContext { + roi, + tgt_client, + bucket, + object, + start_time, + ssec_audit_required, + } = ctx; + let replication_action; + match head_object_for_worker(tgt_client.as_ref(), &tgt_client.bucket, object, roi.version_id.map(|v| v.to_string())).await { + Ok(oi) => { + replication_action = replication_action_for_target_head(&object_info, &oi, roi.op_type); + rinfo.replication_status = ReplicationStatusType::Completed; + if replication_action == ReplicationAction::None { + // An SSE-C replica only counts as converged when the same HEAD + // proves its decryption material survived; a broken ciphertext + // copy from an earlier attempt matches by ETag. + if ssec_audit_required && !settle_ssec_passthrough_evidence(&oi, tgt_client, bucket, object, rinfo).await { + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + return None; + } + if roi.op_type == ReplicationType::ExistingObject + && replication_target_head_is_newer_null_version(&object_info, &oi) + { + warn!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + object = %object, + arn = %tgt_client.arn, + endpoint = %tgt_client.to_url(), + reason = "target_newer_than_source_null_version", + "Skipping replication because newer target version exists" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info.clone(), + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + } + + if object_info.target_replication_status(&tgt_client.arn) == ReplicationStatusType::Pending + || object_info.target_replication_status(&tgt_client.arn) == ReplicationStatusType::Failed + || roi.op_type == ReplicationType::ExistingObject + { + rinfo.replication_action = replication_action; + rinfo.replication_status = ReplicationStatusType::Completed; + } + + if rinfo.replication_status == ReplicationStatusType::Completed + && roi.op_type == ReplicationType::ExistingObject + && !tgt_client.reset_id.is_empty() + { + apply_replication_resync_timestamp(rinfo, &tgt_client.reset_id); + } + + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + + return None; + } + } + Err(e) => { + if is_version_id_format_mismatch(&e) { + // Version-ID format mismatch: retry without versionId and compare ETags. + match head_object_fallback(tgt_client, object).await { + Ok(Some(oi)) => { + replication_action = if replication_etags_match(object_info.etag.as_deref(), oi.e_tag.as_deref()) { + if ssec_audit_required + && !settle_ssec_passthrough_evidence(&oi, tgt_client, bucket, object, rinfo).await + { + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + return None; + } + ReplicationAction::None + } else { + ReplicationAction::All + }; + } + Ok(None) => { + replication_action = ReplicationAction::All; + } + Err(e2) => { + rinfo.error = Some(e2.to_string()); + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + error = %e2, + reason = "head_object_fallback_failed", + "Failed replication head-object fallback" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info, + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + return None; + } + } + } else if e.as_service_error().is_some_and(|se| se.is_not_found()) { + replication_action = ReplicationAction::All; + } else { + rinfo.error = Some(e.to_string()); + debug!( + event = EVENT_RESYNC_RUNTIME_SKIPPED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + error = %e, + reason = "head_object_failed", + "Skipped replication because head-object failed" + ); + + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info, + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); + return None; + } + } + }; + + Some((replication_action, object_info)) +} + +/// Record a fail-closed put-options failure. +/// Unsupported source metadata (e.g. managed SSE) is a fail-closed +/// condition: report FAILED so the composite status and the +/// OperationFailedReplication event reflect that nothing reached +/// the target, instead of leaking the optimistic Completed set earlier. +fn fail_replicate_all_put_options( + rinfo: &mut ReplicatedTargetInfo, + tgt_client: &TargetClient, + bucket: &str, + object_info: ObjectInfo, + e: &Error, + start_time: OffsetDateTime, +) { + rinfo.replication_status = ReplicationStatusType::Failed; + rinfo.error = Some(e.to_string()); + warn!( + event = EVENT_RESYNC_TARGET_OPERATION_FAILED, + component = LOG_COMPONENT_ECSTORE, + subsystem = LOG_SUBSYSTEM_REPLICATION_RESYNC, + bucket = %bucket, + arn = %tgt_client.arn, + operation = "build_put_options", + error = %e, + "Replication target operation failed" + ); + send_local_event(EventArgs { + event_name: EventName::ObjectReplicationNotTracked.to_string(), + bucket_name: bucket.to_string(), + object: object_info, + user_agent: "Internal: [Replication]".to_string(), + ..Default::default() + }); + + rinfo.duration = (OffsetDateTime::now_utc() - start_time).unsigned_abs(); +} + +/// Borrowed inputs shared by both transports of the `replicate_all` payload +/// transfer step. +struct ReplicateAllPayloadContext<'a, S: ReplicationObjectIO> { + storage: &'a Arc, + tgt_client: &'a Arc, + bucket: &'a str, + object: &'a str, + object_info: &'a ObjectInfo, + obj_opts: &'a ObjectOptions, + arn: &'a str, + transfer_size: i64, + is_multipart: bool, + put_opts: PutObjectOptions, +} + +/// Ship the object payload to the replication target over the multipart or +/// single-put transport, returning the transport error when the upload fails. +async fn replicate_all_payload_to_target( + ctx: ReplicateAllPayloadContext<'_, S>, + mut gr: GetObjectReader, +) -> Option { + if ctx.is_multipart { + drop(gr); + let result = replicate_object_with_multipart(MultipartReplicationContext { + storage: ctx.storage.clone(), + cli: ctx.tgt_client.clone(), + src_bucket: ctx.bucket, + dst_bucket: &ctx.tgt_client.bucket, + object: ctx.object, + object_info: ctx.object_info, + obj_opts: ctx.obj_opts, + arn: ctx.arn, + put_opts: ctx.put_opts, + }) + .await; + result.err() + } else { + gr.stream = wrap_with_bandwidth_monitor(gr.stream, &ctx.put_opts, ctx.bucket, ctx.arn); + let byte_stream = async_read_to_bytestream(gr.stream); + let result = ctx + .tgt_client + .put_object(&ctx.tgt_client.bucket, ctx.object, ctx.transfer_size, byte_stream, &ctx.put_opts) + .await + .map(|assigned_version_id| { + audit_target_version_identity( + ctx.tgt_client, + &ctx.put_opts.internal.source_version_id, + assigned_version_id.as_deref(), + ) + }) + .map_err(|e| std::io::Error::other(e.to_string())); + result.err() + } +} + fn wrap_with_bandwidth_monitor_with_header( stream: Box, bucket: &str, @@ -3615,27 +3925,6 @@ mod tests { ReplicationTargetStore::register_test_target(target).await; } - /// P1-19 runtime spot-check exemption matrix: drift only applies when the - /// source addressed a real version uuid. - #[test] - fn test_version_identity_drift_judgment() { - let source = "6fa459ea-ee8a-3ca4-894e-db77e160355e"; - for (sent, got, expected) in [ - (source, Some(source), false), - (source, Some("0e304ce5-33e9-4b8a-9b12-9e40a53e6ded"), true), - (source, None, true), - ("", None, false), - ("null", Some("anything"), false), - ("00000000-0000-0000-0000-000000000000", Some("anything"), false), - ] { - assert_eq!( - version_identity_drifted(sent, got), - expected, - "sent {sent:?} got {got:?} must judge drift = {expected}" - ); - } - } - #[test] fn resync_admission_configuration_is_bounded() { assert_eq!(ENV_REPL_RESYNC_MAX_JOBS, "RUSTFS_REPL_RESYNC_MAX_JOBS"); @@ -3678,15 +3967,6 @@ mod tests { drop((first, second)); } - #[test] - fn replication_target_offline_error_classifier_is_network_scoped() { - assert!(is_replication_target_offline_error(&"put_object dispatch failure: connector error")); - assert!(is_replication_target_offline_error(&"request TimeoutError after retry")); - assert!(is_replication_target_offline_error(&"tcp connect error: connection refused")); - assert!(!is_replication_target_offline_error(&"put_object failed: AccessDenied: denied")); - assert!(!is_replication_target_offline_error(&"put_object failed: NoSuchBucket")); - } - #[tokio::test] async fn replication_target_network_failure_marks_target_offline() { let endpoint = format!("http://network-failure-{}.example:9000", Uuid::new_v4()); @@ -3960,27 +4240,6 @@ mod tests { ); } - /// P1-21 regression guard for the outcome formula. A fully successful - /// delete-marker replication must acknowledge its MRF entry: the formula - /// once carried `&& !requires_delayed_purge`, which pinned every - /// delete-marker entry to Missed and retained the whole backlog forever. - /// (Deterministically staging a marker-creation entry in the durable - /// journal from e2e would require saturating the worker queues, so the - /// formula is pinned here instead; the purge-intent replay half is pinned - /// by the delayed-purge e2e pair.) - #[test] - fn test_replicate_delete_outcome_is_not_held_hostage_by_the_delayed_purge() { - assert!( - replicate_delete_outcome(1, 1, true, true, &ReplicationStatusType::Completed), - "a completed delete-marker replication must be acknowledgeable even though a delayed purge watch is pending" - ); - assert!(!replicate_delete_outcome(0, 0, true, true, &ReplicationStatusType::Completed)); - assert!(!replicate_delete_outcome(2, 1, true, true, &ReplicationStatusType::Completed)); - assert!(!replicate_delete_outcome(1, 1, false, true, &ReplicationStatusType::Completed)); - assert!(!replicate_delete_outcome(1, 1, true, false, &ReplicationStatusType::Completed)); - assert!(!replicate_delete_outcome(1, 1, true, true, &ReplicationStatusType::Failed)); - } - /// P1-21 review follow-up: a target whose recorded marker version is /// inconsistent must be reported as a per-target FAILURE. Treating the /// refusal as success let the watcher and the MRF replay drop the purge @@ -4020,41 +4279,6 @@ mod tests { ); } - #[test] - fn test_delete_marker_purge_mrf_entry_replays_through_the_stale_marker_branch() { - let delete_marker_version_id = Uuid::new_v4(); - let dobj = DeletedObjectReplicationInfo { - delete_object: ReplicationDeletedObject { - object_name: "doc.txt".to_string(), - // A version-purge flavored source event: the entry must still - // be reshaped as a marker-creation delete so replay funnels - // into the stale-marker branch instead of re-running the full - // delete replication (whose source-state stamping would fail - // against the already-purged version). - delete_marker: false, - version_id: Some(Uuid::new_v4()), - delete_marker_version_id: Some(delete_marker_version_id), - ..Default::default() - }, - bucket: "bucket-a".to_string(), - ..Default::default() - }; - - let entry = delete_marker_purge_mrf_entry(&dobj, vec!["arn:a".to_string()]); - - assert!(entry.delete_marker, "purge intents must replay as marker-creation deletes"); - assert_eq!(entry.version_id, None, "the purged data version must not leak into the replay"); - assert_eq!(entry.delete_marker_version_id, Some(delete_marker_version_id)); - assert_eq!( - entry.target_arns, - vec!["arn:a".to_string()], - "only the targets whose purge failed may be retried" - ); - assert_eq!(entry.retry_count, 0); - assert_eq!(entry.bucket, "bucket-a"); - assert_eq!(entry.object, "doc.txt"); - } - #[test] fn test_is_retryable_delete_replication_head_error_allows_delete_marker_head_responses() { assert!( @@ -4562,60 +4786,4 @@ mod tests { assert!(resync_state_accepts_update(¤t, &matching)); assert!(!resync_state_accepts_update(¤t, &stale)); } - - #[test] - fn test_resync_status_duration_only_tracks_terminal_status() { - let start = match OffsetDateTime::from_unix_timestamp(1_700_000_000) { - Ok(start) => start, - Err(err) => panic!("valid test timestamp: {err}"), - }; - let end = start + time::Duration::seconds(2); - - assert_eq!( - resync_status_duration(ResyncStatusType::ResyncCompleted, Some(start), end), - Some(std::time::Duration::from_millis(2000)) - ); - assert_eq!(resync_status_duration(ResyncStatusType::ResyncStarted, Some(start), end), None); - assert_eq!(resync_status_duration(ResyncStatusType::ResyncFailed, None, end), None); - } - - #[test] - fn target_delete_version_id_preserves_explicit_null_purges() { - let version_id = Uuid::new_v4(); - - assert_eq!(target_delete_version_id(version_id, true), Some(version_id.to_string())); - assert_eq!(target_delete_version_id(Uuid::nil(), true).as_deref(), Some(NULL_VERSION_ID)); - assert_eq!(target_delete_version_id(Uuid::nil(), false), None); - } - - #[test] - fn delete_marker_purge_prefers_the_recorded_target_version() { - let source = Uuid::new_v4(); - let arn = "arn:rustfs:replication::target:bucket"; - - // No recorded mapping: fall back to deriving from the source uuid. - assert_eq!(delete_marker_purge_version_id(None, arn, source), Some(Some(source.to_string()))); - - // Recorded mapping wins — a generic S3 target assigns its own id, so the - // derived one would purge the wrong version or nothing at all. - let mut state = ReplicationState::default(); - state - .target_delete_marker_version_ids - .insert(arn.to_string(), "target-assigned-id".to_string()); - assert_eq!( - delete_marker_purge_version_id(Some(&state), arn, source), - Some(Some("target-assigned-id".to_string())) - ); - - // A mapping recorded for a different ARN must not be reused. - assert_eq!( - delete_marker_purge_version_id(Some(&state), "arn:rustfs:replication::other:bucket", source), - Some(Some(source.to_string())) - ); - - // Inconsistent persisted metadata: refuse to purge rather than guess. - let mut corrupt = state.clone(); - corrupt.target_delete_marker_version_ids_corrupt = true; - assert_eq!(delete_marker_purge_version_id(Some(&corrupt), arn, source), None); - } } diff --git a/crates/ecstore/src/bucket/replication/replication_state.rs b/crates/ecstore/src/bucket/replication/replication_state.rs index acd4b696a..ce2f1a36f 100644 --- a/crates/ecstore/src/bucket/replication/replication_state.rs +++ b/crates/ecstore/src/bucket/replication/replication_state.rs @@ -1161,6 +1161,31 @@ mod tests { assert!(all.contains_key("proxy-only-bucket")); } + /// Pins the read-proxy metric contract (backlog#1675 P1-5): the API + /// strings the GET/HEAD/Tagging proxy paths record map onto the + /// get/head/tagging totals, and only unexpected failures raise the + /// failed counters. + #[tokio::test] + async fn test_proxy_stats_map_read_proxy_apis_to_totals() { + let stats = ReplicationStats::new(); + stats.inc_proxy("proxy-bucket", "GetObject", false).await; + stats.inc_proxy("proxy-bucket", "GetObject", true).await; + stats.inc_proxy("proxy-bucket", "HeadObject", false).await; + stats.inc_proxy("proxy-bucket", "GetObjectTagging", false).await; + stats.inc_proxy("proxy-bucket", "PutObjectTagging", false).await; + stats.inc_proxy("proxy-bucket", "DeleteObjectTagging", true).await; + + let metric = stats.get_proxy_stats("proxy-bucket").await; + assert_eq!(metric.get_total, 2); + assert_eq!(metric.get_failed, 1); + assert_eq!(metric.head_total, 1); + assert_eq!(metric.head_failed, 0); + assert_eq!(metric.get_tag_total, 1); + assert_eq!(metric.put_tag_total, 1); + assert_eq!(metric.delete_tag_total, 1); + assert_eq!(metric.delete_tag_failed, 1); + } + #[tokio::test] async fn test_calculate_bucket_replication_stats_merges_resync_metrics() { let stats = ReplicationStats::new(); diff --git a/crates/ecstore/src/bucket/replication/replication_target_boundary.rs b/crates/ecstore/src/bucket/replication/replication_target_boundary.rs index e82ab7e52..87d94779e 100644 --- a/crates/ecstore/src/bucket/replication/replication_target_boundary.rs +++ b/crates/ecstore/src/bucket/replication/replication_target_boundary.rs @@ -36,11 +36,15 @@ use time::OffsetDateTime; use time::format_description::well_known::Rfc3339; pub(crate) use crate::bucket::bucket_target_sys::{ - AdvancedPutOptions, PutObjectOptions, PutObjectPartOptions, RemoveObjectOptions, TargetClient, + AdvancedPutOptions, PutObjectOptions, PutObjectPartOptions, RemoveObjectOptions, TargetClient, resolve_read_api_version_id, }; #[cfg(test)] pub(crate) use crate::bucket::target::BucketTarget; pub(crate) use crate::bucket::target::BucketTargets; +pub use rustfs_replication::SsecPassthroughCapability; +pub(crate) use rustfs_replication::{ + SsecPassthroughGate, is_replication_target_offline_error, ssec_passthrough_gate, version_identity_drifted, +}; use super::replication_config_store::ReplicationConfigStore; use super::replication_error_boundary::{Error, Result}; @@ -65,6 +69,8 @@ static STANDARD_HEADERS: &[&str] = &[ ]; const ERR_REPLICATION_ENCRYPTION_METADATA_UNSUPPORTED: &str = "replication source contains unsupported encryption metadata"; +pub(crate) const ERR_REPLICATION_SSEC_PASSTHROUGH_UNSUPPORTED: &str = "replication target does not support SSE-C passthrough: the replica would lose its decryption material \ + (run ?replication-check to re-probe)"; #[derive(Debug, Clone, Copy, PartialEq, Eq)] enum ReplicationSourceEncryption { @@ -146,6 +152,13 @@ pub(crate) fn replication_object_is_ssec_encrypted(user_defined: &HashMap bool { + rustfs_replication::ssec_passthrough_evidence_present(head.sse_customer_algorithm.as_deref()) +} + pub(crate) struct ReplicationTargetStore; impl ReplicationTargetStore { @@ -165,6 +178,17 @@ impl ReplicationTargetStore { BucketTargetSys::get().mark_target_offline(target_client).await } + /// Returns the cached verdict and whether it has outlived its TTL. + pub(crate) async fn ssec_passthrough_capability(arn: &str) -> (SsecPassthroughCapability, bool) { + BucketTargetSys::get().ssec_passthrough_capability(arn).await + } + + pub(crate) async fn record_ssec_passthrough_capability(arn: &str, capability: SsecPassthroughCapability) { + BucketTargetSys::get() + .record_ssec_passthrough_capability(arn, capability) + .await + } + #[cfg(test)] pub(crate) async fn register_test_target(target_client: &Arc) { BucketTargetSys::get().arn_remotes_map.write().await.insert( @@ -898,6 +922,27 @@ mod tests { } } + /// Pins the HeadObjectOutput field extraction feeding the crate-owned + /// evidence judgment (the gate/evidence policy matrix itself is pinned in + /// `rustfs-replication`'s object tests). + #[test] + fn ssec_passthrough_evidence_requires_customer_algorithm_echo() { + let with_evidence = HeadObjectOutput::builder().sse_customer_algorithm("AES256").build(); + assert!(ssec_passthrough_evidence_present(&with_evidence)); + + let empty_algorithm = HeadObjectOutput::builder().sse_customer_algorithm("").build(); + assert!( + !ssec_passthrough_evidence_present(&empty_algorithm), + "an empty echo is not evidence of preserved SSE-C material" + ); + + let without_evidence = HeadObjectOutput::builder().e_tag("\"abc\"").content_length(8).build(); + assert!( + !ssec_passthrough_evidence_present(&without_evidence), + "a plain HEAD response must classify the target as having dropped the material" + ); + } + #[test] fn replication_put_options_adds_ssec_checksum_metadata() { let metadata = HashMap::from([(SSEC_ALGORITHM_HEADER.to_string(), "AES256".to_string())]); diff --git a/crates/ecstore/src/object_api/types.rs b/crates/ecstore/src/object_api/types.rs index 624cb7593..99ec6c83e 100644 --- a/crates/ecstore/src/object_api/types.rs +++ b/crates/ecstore/src/object_api/types.rs @@ -277,6 +277,20 @@ pub struct ObjectOptions { /// fence avoids recursively acquiring the read lock behind a queued writer. pub bucket_lifecycle_lock_fence: Option, pub replication_request: bool, + /// True when the inbound request carried the + /// `{x-rustfs-,x-minio-}source-proxy-request` header family with the + /// value "true": the request was already proxied by a replication peer, + /// so this server must not proxy a local miss onward (anti-loop, + /// MinIO-compatible). The header only disables proxying — it grants no + /// capability — so no authorization gate is required to honor it. + pub proxy_request: bool, + /// True when the `source-proxy-request` header family was present at + /// all, regardless of value (MinIO's `ProxyHeaderSet`). A replication + /// peer sends `source-proxy-request: false` on its worker convergence + /// HEADs precisely so the receiver answers locally instead of proxying + /// back — otherwise a proxied 404->200 echo makes the worker believe the + /// object already converged and it never replicates it. + pub proxy_header_set: bool, /// Source-cluster LWW timestamps carried by an authorized replication /// request; None when the source never modified the category. Only the /// replication-authorized options builders may set these. diff --git a/crates/filemeta/src/fileinfo.rs b/crates/filemeta/src/fileinfo.rs index 393975c18..42613185e 100644 --- a/crates/filemeta/src/fileinfo.rs +++ b/crates/filemeta/src/fileinfo.rs @@ -42,6 +42,9 @@ const FILEINFO_PART_BITMAP_WORD_BITS: usize = std::mem::size_of::() * 8; const FILEINFO_PART_BITMAP_WORDS: usize = MAX_FILEINFO_PARTS.div_ceil(FILEINFO_PART_BITMAP_WORD_BITS); // Additional constants from Go version +// Intentionally duplicated (S3 wire literal): rustfs-replication and +// rustfs-object-data-cache carry their own independent "null" constants so +// they stay free of a rustfs-filemeta dependency. Keep all three in sync. pub const NULL_VERSION_ID: &str = "null"; // pub const RUSTFS_ERASURE_UPGRADED: &str = "x-rustfs-internal-erasure-upgraded"; diff --git a/crates/replication/src/delete.rs b/crates/replication/src/delete.rs index 3a0c32bf8..3b6fbb016 100644 --- a/crates/replication/src/delete.rs +++ b/crates/replication/src/delete.rs @@ -14,8 +14,13 @@ use std::any::Any; +use uuid::Uuid; + use crate::storage_api::DeletedObject; -use crate::{MrfOpKind, MrfReplicateEntry, ReplicationState, ReplicationType, ReplicationWorkerOperation}; +use crate::{ + MrfOpKind, MrfReplicateEntry, NULL_VERSION_ID, REPLICATE_EXISTING_DELETE, ReplicateObjectInfo, ReplicationState, + ReplicationStatusType, ReplicationType, ReplicationWorkerOperation, +}; #[derive(Debug, Clone, Default)] pub struct DeletedObjectReplicationInfo { @@ -117,17 +122,114 @@ pub fn is_retryable_delete_replication_head_error(is_not_found: bool, code: Opti !(is_not_found || matches!(code, Some("MethodNotAllowed" | "405"))) } +/// Build the delete-replication work item for an existing delete marker or +/// version purge discovered during a resync scan. +pub fn resync_existing_delete_replication_info(roi: &ReplicateObjectInfo, target_arn: &str) -> DeletedObjectReplicationInfo { + let (version_id, dm_version_id) = if roi.version_purge_status.is_empty() { + (None, roi.version_id) + } else { + (roi.version_id, None) + }; + + DeletedObjectReplicationInfo { + delete_object: DeletedObject { + object_name: roi.name.clone(), + delete_marker_version_id: dm_version_id, + version_id, + replication_state: roi.replication_state.clone(), + delete_marker: roi.delete_marker, + delete_marker_mtime: roi.mod_time, + ..Default::default() + }, + bucket: roi.bucket.clone(), + event_type: REPLICATE_EXISTING_DELETE.to_string(), + op_type: ReplicationType::ExistingObject, + target_arn: target_arn.to_string(), + ..Default::default() + } +} + +/// Whether a delete replication fully succeeded — the MRF replay acknowledges +/// (drops) an entry exactly when this returns true. +/// +/// The delayed purge is deliberately NOT an input: holding the outcome hostage +/// to it (`&& !requires_delayed_purge`) forced `false` for every delete-marker +/// entry and retained them all in the durable MRF journal forever. Purge +/// failures persist their own purge-intent entry instead +/// (`watch_and_purge_source_delete_marker`), and replays of those entries +/// report purge success through `purge_stale_delete_marker_targets`. +pub fn replicate_delete_outcome( + expected_targets: usize, + replicated_targets: usize, + state_persisted: bool, + source_state_verified: bool, + replication_status: &ReplicationStatusType, +) -> bool { + expected_targets > 0 + && replicated_targets == expected_targets + && state_persisted + && source_state_verified + && *replication_status == ReplicationStatusType::Completed +} + +pub fn target_delete_version_id(version_id: Uuid, version_purge: bool) -> Option { + if version_id.is_nil() { + version_purge.then(|| NULL_VERSION_ID.to_string()) + } else { + Some(version_id.to_string()) + } +} + +/// Which version a delete-marker purge should address on one target. +/// +/// `None` means do not purge at all: the recorded mapping disagreed across the +/// dual internal prefixes, and guessing an id could destroy a live version on +/// the target. `Some(id)` is the exact version the target reported when it +/// accepted the marker; falling back to a source-derived id is only correct +/// when the target mirrors source version ids, which a generic S3 target does +/// not. +pub fn delete_marker_purge_version_id( + state: Option<&ReplicationState>, + arn: &str, + delete_marker_version_id: Uuid, +) -> Option> { + if state.is_some_and(|state| state.target_delete_marker_version_ids_corrupt) { + return None; + } + let recorded = state.and_then(|state| state.target_delete_marker_version_ids.get(arn).cloned()); + Some(match recorded { + Some(version_id) => Some(version_id), + None => target_delete_version_id(delete_marker_version_id, true), + }) +} + +/// Shape an exhausted purge intent as a marker-creation delete entry. Replay +/// reconstructs it with `delete_marker: true`, finds the source marker gone, +/// and funnels into the stale-marker branch of `replicate_delete_with_outcome` +/// — which re-runs the purge without touching source state and reports purge +/// success as the replay outcome. +pub fn delete_marker_purge_mrf_entry(dobj: &DeletedObjectReplicationInfo, failed_arns: Vec) -> MrfReplicateEntry { + let mut entry = dobj.to_mrf_entry(); + entry.delete_marker = true; + entry.version_id = None; + entry.retry_count = 0; + entry.target_arns = failed_arns; + entry +} + #[cfg(test)] mod tests { use std::collections::HashMap; use super::{ - DeletedObjectReplicationInfo, is_retryable_delete_replication_head_error, is_version_delete_replication, - should_retry_delete_marker_purge, + DeletedObjectReplicationInfo, delete_marker_purge_mrf_entry, delete_marker_purge_version_id, + is_retryable_delete_replication_head_error, is_version_delete_replication, replicate_delete_outcome, + should_retry_delete_marker_purge, target_delete_version_id, }; use crate::storage_api::DeletedObject; use crate::{ - MrfOpKind, ReplicationState, ReplicationStatusType, ReplicationType, ReplicationWorkerOperation, VersionPurgeStatusType, + MrfOpKind, NULL_VERSION_ID, ReplicationState, ReplicationStatusType, ReplicationType, ReplicationWorkerOperation, + VersionPurgeStatusType, }; use uuid::Uuid; @@ -328,4 +430,100 @@ mod tests { assert!(!is_retryable_delete_replication_head_error(true, Some("NoSuchKey"))); assert!(is_retryable_delete_replication_head_error(false, Some("AccessDenied"))); } + + /// P1-21 regression guard for the outcome formula. A fully successful + /// delete-marker replication must acknowledge its MRF entry: the formula + /// once carried `&& !requires_delayed_purge`, which pinned every + /// delete-marker entry to Missed and retained the whole backlog forever. + /// (Deterministically staging a marker-creation entry in the durable + /// journal from e2e would require saturating the worker queues, so the + /// formula is pinned here instead; the purge-intent replay half is pinned + /// by the delayed-purge e2e pair.) + #[test] + fn test_replicate_delete_outcome_is_not_held_hostage_by_the_delayed_purge() { + assert!( + replicate_delete_outcome(1, 1, true, true, &ReplicationStatusType::Completed), + "a completed delete-marker replication must be acknowledgeable even though a delayed purge watch is pending" + ); + assert!(!replicate_delete_outcome(0, 0, true, true, &ReplicationStatusType::Completed)); + assert!(!replicate_delete_outcome(2, 1, true, true, &ReplicationStatusType::Completed)); + assert!(!replicate_delete_outcome(1, 1, false, true, &ReplicationStatusType::Completed)); + assert!(!replicate_delete_outcome(1, 1, true, false, &ReplicationStatusType::Completed)); + assert!(!replicate_delete_outcome(1, 1, true, true, &ReplicationStatusType::Failed)); + } + + #[test] + fn test_delete_marker_purge_mrf_entry_replays_through_the_stale_marker_branch() { + let delete_marker_version_id = Uuid::new_v4(); + let dobj = DeletedObjectReplicationInfo { + delete_object: DeletedObject { + object_name: "doc.txt".to_string(), + // A version-purge flavored source event: the entry must still + // be reshaped as a marker-creation delete so replay funnels + // into the stale-marker branch instead of re-running the full + // delete replication (whose source-state stamping would fail + // against the already-purged version). + delete_marker: false, + version_id: Some(Uuid::new_v4()), + delete_marker_version_id: Some(delete_marker_version_id), + ..Default::default() + }, + bucket: "bucket-a".to_string(), + ..Default::default() + }; + + let entry = delete_marker_purge_mrf_entry(&dobj, vec!["arn:a".to_string()]); + + assert!(entry.delete_marker, "purge intents must replay as marker-creation deletes"); + assert_eq!(entry.version_id, None, "the purged data version must not leak into the replay"); + assert_eq!(entry.delete_marker_version_id, Some(delete_marker_version_id)); + assert_eq!( + entry.target_arns, + vec!["arn:a".to_string()], + "only the targets whose purge failed may be retried" + ); + assert_eq!(entry.retry_count, 0); + assert_eq!(entry.bucket, "bucket-a"); + assert_eq!(entry.object, "doc.txt"); + } + + #[test] + fn target_delete_version_id_preserves_explicit_null_purges() { + let version_id = Uuid::new_v4(); + + assert_eq!(target_delete_version_id(version_id, true), Some(version_id.to_string())); + assert_eq!(target_delete_version_id(Uuid::nil(), true).as_deref(), Some(NULL_VERSION_ID)); + assert_eq!(target_delete_version_id(Uuid::nil(), false), None); + } + + #[test] + fn delete_marker_purge_prefers_the_recorded_target_version() { + let source = Uuid::new_v4(); + let arn = "arn:rustfs:replication::target:bucket"; + + // No recorded mapping: fall back to deriving from the source uuid. + assert_eq!(delete_marker_purge_version_id(None, arn, source), Some(Some(source.to_string()))); + + // Recorded mapping wins — a generic S3 target assigns its own id, so the + // derived one would purge the wrong version or nothing at all. + let mut state = ReplicationState::default(); + state + .target_delete_marker_version_ids + .insert(arn.to_string(), "target-assigned-id".to_string()); + assert_eq!( + delete_marker_purge_version_id(Some(&state), arn, source), + Some(Some("target-assigned-id".to_string())) + ); + + // A mapping recorded for a different ARN must not be reused. + assert_eq!( + delete_marker_purge_version_id(Some(&state), "arn:rustfs:replication::other:bucket", source), + Some(Some(source.to_string())) + ); + + // Inconsistent persisted metadata: refuse to purge rather than guess. + let mut corrupt = state.clone(); + corrupt.target_delete_marker_version_ids_corrupt = true; + assert_eq!(delete_marker_purge_version_id(Some(&corrupt), arn, source), None); + } } diff --git a/crates/replication/src/filemeta.rs b/crates/replication/src/filemeta.rs index 68064dc19..8dfdb26c5 100644 --- a/crates/replication/src/filemeta.rs +++ b/crates/replication/src/filemeta.rs @@ -27,6 +27,11 @@ use uuid::Uuid; pub const REPLICATION_RESET: &str = "replication-reset"; pub const REPLICATION_STATUS: &str = "replication-status"; +/// The S3 wire spelling of the unversioned ("null") version id. Owned here as +/// part of the replication wire contracts; `rustfs-filemeta` keeps its own +/// copy of the same literal (the crates are intentionally independent). +pub const NULL_VERSION_ID: &str = "null"; + // ReplicateQueued - replication being queued trail pub const REPLICATE_QUEUED: &str = "replicate:queue"; diff --git a/crates/replication/src/lib.rs b/crates/replication/src/lib.rs index bb0c08385..cb0761b5c 100644 --- a/crates/replication/src/lib.rs +++ b/crates/replication/src/lib.rs @@ -37,11 +37,12 @@ pub use config::{ validate_replication_config_target_arns, }; pub use delete::{ - DeletedObjectReplicationInfo, is_retryable_delete_replication_head_error, is_version_delete_replication, - should_retry_delete_marker_purge, + DeletedObjectReplicationInfo, delete_marker_purge_mrf_entry, delete_marker_purge_version_id, + is_retryable_delete_replication_head_error, is_version_delete_replication, replicate_delete_outcome, + resync_existing_delete_replication_info, should_retry_delete_marker_purge, target_delete_version_id, }; pub use filemeta::{ - REPLICATE_EXISTING, REPLICATE_EXISTING_DELETE, REPLICATE_HEAL, REPLICATE_HEAL_DELETE, REPLICATE_INCOMING, + NULL_VERSION_ID, REPLICATE_EXISTING, REPLICATE_EXISTING_DELETE, REPLICATE_HEAL, REPLICATE_HEAL_DELETE, REPLICATE_INCOMING, REPLICATE_INCOMING_DELETE, REPLICATE_MRF, REPLICATE_QUEUED, REPLICATION_RESET, REPLICATION_STATUS, ReplicateDecision, ReplicateObjectInfo, ReplicateTargetDecision, ReplicatedInfos, ReplicatedTargetInfo, ReplicationAction, ReplicationState, ReplicationStatusType, ReplicationType, ReplicationWorkerOperation, ResyncDecision, ResyncTargetDecision, @@ -58,8 +59,9 @@ pub use multipart::{ replication_multipart_complete_actual_size, replication_multipart_part_plan, }; pub use object::{ - ReplicationSourceObject, ReplicationTargetObject, content_matches_by_etag, replication_action_for_target, - replication_etags_match, target_is_newer_than_source_null_version, + ReplicationSourceObject, ReplicationTargetObject, SsecPassthroughCapability, SsecPassthroughGate, content_matches_by_etag, + is_replication_target_offline_error, replication_action_for_target, replication_etags_match, + ssec_passthrough_evidence_present, ssec_passthrough_gate, target_is_newer_than_source_null_version, version_identity_drifted, }; pub use operation::{ MustReplicateOptions, ReplicationDeleteScheduleInput, ReplicationDeleteSource, ReplicationDeleteStateSource, @@ -76,7 +78,7 @@ pub use queue::{ pub use resync::{ BucketReplicationResyncStatus, Error, RESYNC_FILE_MAX_BYTES, Result, ResyncOpts, ResyncStatusType, TargetReplicationResyncStatus, decode_resync_file, encode_resync_file, is_version_id_mismatch, resync_state_accepts_update, - sanitize_resync_error_detail, should_auto_resume_resync, should_count_head_proxy_failure, + resync_status_duration, sanitize_resync_error_detail, should_auto_resume_resync, should_count_head_proxy_failure, }; pub use rule::ReplicationRuleExt; pub use runtime::{ diff --git a/crates/replication/src/object.rs b/crates/replication/src/object.rs index afbb2ef46..7ce7761ca 100644 --- a/crates/replication/src/object.rs +++ b/crates/replication/src/object.rs @@ -157,11 +157,130 @@ fn comparable_metadata(metadata: Option<&HashMap>) -> HashMap) -> bool { + if source_version_id.is_empty() { + return false; + } + // A nil source uuid travels as the literal "null" (unversioned-source + // semantics); no identity contract applies to it. + if uuid::Uuid::parse_str(source_version_id) + .map(|uuid| uuid.is_nil()) + .unwrap_or(true) + { + return false; + } + assigned_version_id != Some(source_version_id) +} + +const REPLICATION_TARGET_OFFLINE_ERROR_MARKERS: &[&str] = &[ + "dispatch failure", + "timeouterror", + "timed out", + "connection refused", + "connection reset", + "connection closed", + "connection aborted", + "broken pipe", + "dns error", + "failed to lookup address", + "name or service not known", + "deadline has elapsed", + "tcp connect error", +]; + +/// True when a target operation error reads as a network/transport failure — +/// the only class of error that should mark a replication target offline. +pub fn is_replication_target_offline_error(err: &(impl std::fmt::Display + ?Sized)) -> bool { + let message = err.to_string().to_ascii_lowercase(); + REPLICATION_TARGET_OFFLINE_ERROR_MARKERS + .iter() + .any(|marker| message.contains(marker)) +} + +/// Whether a replication target preserves the SSE-C passthrough transport +/// headers (`X-Rustfs-Replication-*`) end to end. +/// +/// A target that silently drops those headers (MinIO, generic S3) stores the +/// forwarded ciphertext without its decryption material — an unreadable +/// replica that used to report COMPLETED. The replication worker audits the +/// first passthrough PUT per target (HEAD-back for SSE-C evidence) and caches +/// the verdict; a fresh `Unsupported` fails SSE-C replication closed before +/// any PUT is sent. The verdict cache (per-ARN map, lifecycle, and TTL) is +/// owned by the runtime's bucket target system; this crate owns only the +/// verdict vocabulary and the gate policy below. +#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)] +pub enum SsecPassthroughCapability { + #[default] + Unknown, + Supported, + Unsupported, +} + +/// Fail-closed decision for an SSE-C passthrough replication attempt, derived +/// from the target's cached [`SsecPassthroughCapability`]. Pure so the policy +/// can migrate with the worker (M2) without dragging the cache along; the +/// caller computes `expired` from the cache record's age (see the runtime's +/// `SSEC_PASSTHROUGH_CAPABILITY_TTL`). +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum SsecPassthroughGate { + /// Not an SSE-C object, or the target has a fresh proof that it preserves + /// the passthrough transport headers: replicate without a HEAD-back audit. + Proceed, + /// No usable verdict — first SSE-C attempt since the target was (re)built, + /// or the recorded verdict (in either direction) aged out: PUT, then HEAD + /// the replica back and require SSE-C evidence before reporting COMPLETED. + ProceedWithAudit, + /// The target was recently proven to drop the passthrough headers: do not + /// send the PUT, report FAILED (the object stays on the normal MRF retry + /// channel and re-audits once the verdict expires). + FailClosed, +} + +pub fn ssec_passthrough_gate(ssec: bool, capability: SsecPassthroughCapability, expired: bool) -> SsecPassthroughGate { + if !ssec { + return SsecPassthroughGate::Proceed; + } + // An expired verdict — Supported or Unsupported — must be re-earned: a + // stale Unsupported would otherwise stick forever after a target upgrade, + // and a stale Supported would fail open after a backend swap behind the + // same endpoint. + if expired { + return SsecPassthroughGate::ProceedWithAudit; + } + match capability { + SsecPassthroughCapability::Supported => SsecPassthroughGate::Proceed, + SsecPassthroughCapability::Unknown => SsecPassthroughGate::ProceedWithAudit, + SsecPassthroughCapability::Unsupported => SsecPassthroughGate::FailClosed, + } +} + +/// True when a replication-check HEAD of the replica proves the SSE-C +/// material survived passthrough: a RustFS target restores the transport +/// headers into the stored SSE-C keys and its HEAD echoes +/// `x-amz-server-side-encryption-customer-algorithm` (the replication-check +/// exemption skips key validation but not the metadata echo). A target that +/// dropped the headers stored a plain object and echoes nothing. The caller +/// extracts the echoed customer-algorithm value from its HEAD response type. +pub fn ssec_passthrough_evidence_present(sse_customer_algorithm: Option<&str>) -> bool { + sse_customer_algorithm.is_some_and(|algo| !algo.is_empty()) +} + #[cfg(test)] mod tests { use super::{ - ReplicationSourceObject, ReplicationTargetObject, content_matches_by_etag, replication_action_for_target, - replication_etags_match, target_is_newer_than_source_null_version, + ReplicationSourceObject, ReplicationTargetObject, SsecPassthroughCapability, SsecPassthroughGate, + content_matches_by_etag, is_replication_target_offline_error, replication_action_for_target, replication_etags_match, + ssec_passthrough_evidence_present, ssec_passthrough_gate, target_is_newer_than_source_null_version, + version_identity_drifted, }; use crate::filemeta::{ReplicationAction, ReplicationType}; use crate::http::AMZ_OBJECT_LOCK_MODE; @@ -269,6 +388,96 @@ mod tests { ); } + /// P1-19 runtime spot-check exemption matrix: drift only applies when the + /// source addressed a real version uuid. + #[test] + fn test_version_identity_drift_judgment() { + let source = "6fa459ea-ee8a-3ca4-894e-db77e160355e"; + for (sent, got, expected) in [ + (source, Some(source), false), + (source, Some("0e304ce5-33e9-4b8a-9b12-9e40a53e6ded"), true), + (source, None, true), + ("", None, false), + ("null", Some("anything"), false), + ("00000000-0000-0000-0000-000000000000", Some("anything"), false), + ] { + assert_eq!( + version_identity_drifted(sent, got), + expected, + "sent {sent:?} got {got:?} must judge drift = {expected}" + ); + } + } + + #[test] + fn replication_target_offline_error_classifier_is_network_scoped() { + assert!(is_replication_target_offline_error("put_object dispatch failure: connector error")); + assert!(is_replication_target_offline_error("request TimeoutError after retry")); + assert!(is_replication_target_offline_error("tcp connect error: connection refused")); + assert!(!is_replication_target_offline_error("put_object failed: AccessDenied: denied")); + assert!(!is_replication_target_offline_error("put_object failed: NoSuchBucket")); + } + + /// N2 fail-closed policy: SSE-C replication may only proceed silently + /// against a target with a FRESH proof that it preserves the passthrough + /// transport headers. Unknown targets must be audited; freshly-flagged + /// dropping targets must never receive the PUT; an expired verdict in + /// EITHER direction must be re-earned through the audit — a sticky + /// Unsupported would outlive a target upgrade, and a sticky Supported + /// would fail open after a backend swap behind the same endpoint. + #[test] + fn ssec_passthrough_gate_is_fail_closed_and_ttl_bounded() { + for capability in [ + SsecPassthroughCapability::Unknown, + SsecPassthroughCapability::Supported, + SsecPassthroughCapability::Unsupported, + ] { + for expired in [false, true] { + assert_eq!( + ssec_passthrough_gate(false, capability, expired), + SsecPassthroughGate::Proceed, + "non-SSE-C objects must never be gated on the passthrough capability" + ); + } + } + assert_eq!( + ssec_passthrough_gate(true, SsecPassthroughCapability::Supported, false), + SsecPassthroughGate::Proceed + ); + assert_eq!( + ssec_passthrough_gate(true, SsecPassthroughCapability::Unknown, false), + SsecPassthroughGate::ProceedWithAudit + ); + assert_eq!( + ssec_passthrough_gate(true, SsecPassthroughCapability::Unsupported, false), + SsecPassthroughGate::FailClosed + ); + // Expiry flips both directions back to the audit. + assert_eq!( + ssec_passthrough_gate(true, SsecPassthroughCapability::Unsupported, true), + SsecPassthroughGate::ProceedWithAudit, + "an expired Unsupported verdict must allow a re-audit (upgraded target recovers without operator action)" + ); + assert_eq!( + ssec_passthrough_gate(true, SsecPassthroughCapability::Supported, true), + SsecPassthroughGate::ProceedWithAudit, + "an expired Supported verdict must be re-proven (backend swap behind the same endpoint must not fail open)" + ); + } + + #[test] + fn ssec_passthrough_evidence_requires_customer_algorithm_echo() { + assert!(ssec_passthrough_evidence_present(Some("AES256"))); + assert!( + !ssec_passthrough_evidence_present(Some("")), + "an empty echo is not evidence of preserved SSE-C material" + ); + assert!( + !ssec_passthrough_evidence_present(None), + "a plain HEAD response must classify the target as having dropped the material" + ); + } + #[test] fn replication_action_detects_tags_and_object_lock_metadata_differences() { let mut source_metadata = HashMap::new(); diff --git a/crates/replication/src/resync.rs b/crates/replication/src/resync.rs index 2380d9e18..144181dd2 100644 --- a/crates/replication/src/resync.rs +++ b/crates/replication/src/resync.rs @@ -309,6 +309,31 @@ pub fn is_version_id_mismatch(code: Option<&str>, raw_status: Option) -> bo } } +pub fn resync_status_duration( + status: ResyncStatusType, + start_time: Option, + now: OffsetDateTime, +) -> Option { + if !matches!( + status, + ResyncStatusType::ResyncCompleted | ResyncStatusType::ResyncFailed | ResyncStatusType::ResyncCanceled + ) { + return None; + } + + let millis = (now - start_time?).whole_milliseconds(); + if millis < 0 { + return None; + } + + let millis = if millis > i128::from(u64::MAX) { + u64::MAX + } else { + u64::try_from(millis).ok()? + }; + Some(std::time::Duration::from_millis(millis)) +} + #[derive(Debug, Clone, Serialize, Deserialize, Default)] pub struct BucketReplicationResyncStatus { pub version: u16, @@ -712,6 +737,22 @@ mod tests { assert!(!should_auto_resume_resync(ResyncStatusType::ResyncFailed)); } + #[test] + fn test_resync_status_duration_only_tracks_terminal_status() { + let start = match OffsetDateTime::from_unix_timestamp(1_700_000_000) { + Ok(start) => start, + Err(err) => panic!("valid test timestamp: {err}"), + }; + let end = start + time::Duration::seconds(2); + + assert_eq!( + resync_status_duration(ResyncStatusType::ResyncCompleted, Some(start), end), + Some(std::time::Duration::from_millis(2000)) + ); + assert_eq!(resync_status_duration(ResyncStatusType::ResyncStarted, Some(start), end), None); + assert_eq!(resync_status_duration(ResyncStatusType::ResyncFailed, None, end), None); + } + #[test] fn resync_state_accepts_update_only_for_matching_run() { let current = TargetReplicationResyncStatus { diff --git a/rustfs/src/admin/handlers/site_replication.rs b/rustfs/src/admin/handlers/site_replication.rs index 18eef001b..219ebbffe 100644 --- a/rustfs/src/admin/handlers/site_replication.rs +++ b/rustfs/src/admin/handlers/site_replication.rs @@ -66,18 +66,20 @@ use rustfs_config::{ }; use rustfs_iam::error::is_err_no_such_service_account; use rustfs_iam::federation::OIDC_VIRTUAL_PARENT_CLAIM; +use rustfs_iam::store::object::ObjectStore; use rustfs_iam::store::{MappedPolicy, UserType, sr_wire_user_type, user_type_from_sr_wire}; use rustfs_iam::sys::{ - NewServiceAccountOpts, SITE_REPLICATOR_SERVICE_ACCOUNT, UpdateServiceAccountOpts, get_claims_from_token_with_secret, + IamSys, NewServiceAccountOpts, SITE_REPLICATOR_SERVICE_ACCOUNT, UpdateServiceAccountOpts, get_claims_from_token_with_secret, }; use rustfs_madmin::{ AddOrUpdateUserReq, BucketBandwidth, GroupAddRemove, GroupStatus, IDPSettings, InProgressMetric, InQueueMetric, LDAPConfigSettings, LDAPSettings, OpenIDProviderSettings, PeerInfo, PeerSite, QStat, ReplProxyMetric, ReplicateAddStatus, ReplicateEditStatus, ReplicateRemoveStatus, ResyncBucketStatus, SITE_REPL_API_VERSION, SR_IAM_ITEM_STS_ACC, SR_IAM_ITEM_STS_ACC_LEGACY, SRBucketInfo, SRBucketMeta, SRBucketStatsSummary, SRGroupInfo, SRGroupStatsSummary, SRIAMItem, - SRIAMPolicy, SRILMExpiryStatsSummary, SRInfo, SRMetric, SRMetricsSummary, SRPeerError, SRPeerJoinReq, SRPendingOperation, - SRPolicyMapping, SRPolicyStatsSummary, SRRemoveReq, SRResyncOpStatus, SRRetryStats, SRSessionPolicy, SRSiteSummary, - SRStateEditReq, SRStateInfo, SRStatusInfo, SRSvcAccCreate, SRUserStatsSummary, SiteReplicationInfo, SyncStatus, WorkerStat, + SRIAMPolicy, SRIAMUser, SRILMExpiryStatsSummary, SRInfo, SRMetric, SRMetricsSummary, SRPeerError, SRPeerJoinReq, + SRPendingOperation, SRPolicyMapping, SRPolicyStatsSummary, SRRemoveReq, SRResyncOpStatus, SRRetryStats, SRSTSCredential, + SRSessionPolicy, SRSiteSummary, SRStateEditReq, SRStateInfo, SRStatusInfo, SRSvcAccChange, SRSvcAccCreate, + SRUserStatsSummary, SiteReplicationInfo, SyncStatus, WorkerStat, }; use rustfs_policy::policy::{ Policy, @@ -4584,7 +4586,10 @@ async fn build_metrics_summary(local_peer: &PeerInfo) -> SRMetricsSummary { head_failed_total: non_negative_u64(node.proxy_head_failed), put_tag_total: non_negative_u64(node.proxy_put_tag_total), put_tag_failed_total: non_negative_u64(node.proxy_put_tag_failed), - ..Default::default() + get_tag_total: non_negative_u64(node.proxy_get_tag_total), + get_tag_failed_total: non_negative_u64(node.proxy_get_tag_failed), + remove_tag_total: non_negative_u64(node.proxy_delete_tag_total), + remove_tag_failed_total: non_negative_u64(node.proxy_delete_tag_failed), }, metrics, uptime: node.uptime, @@ -9358,247 +9363,16 @@ async fn apply_iam_item(item: SRIAMItem) -> S3Result<()> { let incoming_updated_at = item.updated_at; match item.r#type.as_str() { - "policy" => { - if let Some(policy) = item.policy { - let policy: Policy = - serde_json::from_value(policy).map_err(|e| s3_error!(InvalidRequest, "invalid policy body: {}", e))?; - iam_sys.set_policy(&item.name, policy).await.map_err(ApiError::from)?; - } else { - iam_sys.delete_policy(&item.name, true).await.map_err(ApiError::from)?; - } - Ok(()) - } - "policy-mapping" => { - let Some(mapping) = item.policy_mapping else { - return Err(s3_error!(InvalidRequest, "policyMapping is required")); - }; - let user_type = - user_type_from_sr_wire(mapping.user_type).ok_or_else(|| s3_error!(InvalidRequest, "invalid userType"))?; - iam_sys - .policy_db_set(&mapping.user_or_group, user_type, mapping.is_group, &mapping.policy) - .await - .map_err(ApiError::from)?; - Ok(()) - } - "group-info" => { - let Some(group_info) = item.group_info else { - return Err(s3_error!(InvalidRequest, "groupInfo is required")); - }; - let update = group_info.update_req; - if !group_info_requires_upsert(&update) { - iam_sys - .remove_users_from_group(&update.group, update.members) - .await - .map_err(ApiError::from)?; - return Ok(()); - } - - iam_sys - .add_users_to_group(&update.group, update.members) - .await - .map_err(ApiError::from)?; - iam_sys - .set_group_status(&update.group, matches!(update.status, GroupStatus::Enabled)) - .await - .map_err(ApiError::from)?; - Ok(()) - } + "policy" => apply_iam_policy_item(&iam_sys, &item.name, item.policy).await, + "policy-mapping" => apply_iam_policy_mapping_item(&iam_sys, item.policy_mapping).await, + "group-info" => apply_iam_group_info_item(&iam_sys, item.group_info).await, // MinIO madmin-go sends `SRIAMItemSTSAcc = "sts-account"`. The legacy alias // `sts-credential` (emitted by older RustFS releases) stays accepted permanently // so mixed-version RustFS sites keep replicating STS credentials during rolling // upgrades; it is a compatibility layer, not temporary code. - SR_IAM_ITEM_STS_ACC | SR_IAM_ITEM_STS_ACC_LEGACY => { - let Some(sts_credential) = item.sts_credential else { - return Err(s3_error!(InvalidRequest, "stsCredential is required")); - }; - let Some(secret) = current_token_signing_key() else { - return Err(s3_error!(InvalidRequest, "token signing key not initialized")); - }; - let claims = get_claims_from_token_with_secret(&sts_credential.session_token, &secret) - .map_err(|e| s3_error!(InvalidRequest, "invalid STS session token: {e}"))?; - let expiration = claims - .get("exp") - .and_then(claims_unix_timestamp) - .map(OffsetDateTime::from_unix_timestamp) - .transpose() - .map_err(|e| s3_error!(InvalidRequest, "invalid STS expiry: {e}"))?; - let groups = string_list_claim(&claims, "groups"); - let compatibility_policy = sts_replication_compatibility_policy(&claims, &sts_credential.parent_policy_mapping); - let cred = rustfs_credentials::Credentials { - access_key: sts_credential.access_key.clone(), - secret_key: sts_credential.secret_key.clone(), - session_token: sts_credential.session_token.clone(), - expiration, - status: "on".to_string(), - parent_user: sts_credential.parent_user.clone(), - groups, - claims: Some(claims), - ..Default::default() - }; - iam_sys - .set_temp_user(&sts_credential.access_key, &cred, compatibility_policy) - .await - .map_err(ApiError::from)?; - Ok(()) - } - "iam-user" => { - let Some(user) = item.iam_user else { - return Err(s3_error!(InvalidRequest, "iamUser is required")); - }; - if let Some(local) = iam_sys.get_user(&user.access_key).await - && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) - { - return Ok(()); - } - if user.is_delete_req { - iam_sys.delete_user(&user.access_key, true).await.map_err(ApiError::from)?; - } else { - let Some(user_req) = user.user_req else { - return Err(s3_error!(InvalidRequest, "userReq is required")); - }; - let is_status_only_update = user_req.secret_key.is_empty() && user_req.policy.is_none(); - if is_status_only_update { - iam_sys - .set_user_status(&user.access_key, user_req.status) - .await - .map_err(ApiError::from)?; - } else { - iam_sys - .create_user(&user.access_key, &user_req) - .await - .map_err(ApiError::from)?; - } - } - Ok(()) - } - "service-account" => { - let Some(change) = item.svc_acc_change else { - return Err(s3_error!(InvalidRequest, "serviceAccountChange is required")); - }; - let envelope = change.oidc_service_account_envelope; - if let Some(create) = change.create { - let local_updated_at = iam_sys - .get_user(&create.access_key) - .await - .map(|local| local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH)); - let replicated_policy = if create.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT { - if local_updated_at.is_some_and(|local_updated_at| is_stale_update(local_updated_at, incoming_updated_at)) { - return Ok(()); - } - ReplicatedServiceAccountPolicy { - policy: Some(site_replicator_service_account_policy()?), - is_envelope: false, - } - } else { - let Some(replicated_policy) = decode_service_account_replication_policy( - &create, - envelope.as_ref(), - incoming_updated_at, - local_updated_at, - )? - else { - return Ok(()); - }; - replicated_policy - }; - match iam_sys.get_service_account(&create.access_key).await { - Ok((existing, _)) => { - if existing.parent_user != create.parent { - return Err(s3_error!( - InvalidRequest, - "service account {} already exists with a different parent user", - create.access_key - )); - } - iam_sys - .update_service_account( - &create.access_key, - UpdateServiceAccountOpts { - name: replicated_policy.metadata_for_existing_account(create.name), - description: replicated_policy.metadata_for_existing_account(create.description), - session_policy: replicated_policy.for_existing_account(), - secret_key: Some(create.secret_key), - expiration: create.expiration, - status: (!create.status.is_empty()).then_some(create.status), - parent_user: None, - allow_site_replicator_account: create.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT, - }, - ) - .await - .map_err(ApiError::from)?; - } - Err(err) if is_err_no_such_service_account(&err) => { - iam_sys - .new_service_account( - &create.parent, - Some(create.groups), - NewServiceAccountOpts { - session_policy: replicated_policy.policy, - access_key: create.access_key, - secret_key: create.secret_key, - name: (!create.name.is_empty()).then_some(create.name), - description: (!create.description.is_empty()).then_some(create.description), - expiration: create.expiration, - allow_site_replicator_account: true, - claims: Some(create.claims), - }, - ) - .await - .map_err(ApiError::from)?; - } - Err(err) => return Err(ApiError::from(err).into()), - } - return Ok(()); - } - - if let Some(update) = change.update { - if let Some(local) = iam_sys.get_user(&update.access_key).await - && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) - { - return Ok(()); - } - let allow_site_replicator_account = update.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT; - let session_policy = if allow_site_replicator_account { - Some(site_replicator_service_account_policy()?) - } else { - update.session_policy.as_str().and_then(|raw| serde_json::from_str(raw).ok()) - }; - iam_sys - .update_service_account( - &update.access_key, - UpdateServiceAccountOpts { - session_policy, - secret_key: (!update.secret_key.is_empty()).then_some(update.secret_key), - name: (!update.name.is_empty()).then_some(update.name), - description: (!update.description.is_empty()).then_some(update.description), - expiration: update.expiration, - status: (!update.status.is_empty()).then_some(update.status), - // Peers replicate credentials, never the local parent binding: - // each site resolves its own parent from its own IAM. - parent_user: None, - allow_site_replicator_account, - }, - ) - .await - .map_err(ApiError::from)?; - return Ok(()); - } - - if let Some(delete) = change.delete { - if let Some(local) = iam_sys.get_user(&delete.access_key).await - && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) - { - return Ok(()); - } - iam_sys - .delete_service_account(&delete.access_key, true) - .await - .map_err(ApiError::from)?; - return Ok(()); - } - - Err(s3_error!(InvalidRequest, "serviceAccountChange is empty")) - } + SR_IAM_ITEM_STS_ACC | SR_IAM_ITEM_STS_ACC_LEGACY => apply_iam_sts_account_item(&iam_sys, item.sts_credential).await, + "iam-user" => apply_iam_user_item(&iam_sys, item.iam_user, incoming_updated_at).await, + "service-account" => apply_iam_service_account_item(&iam_sys, item.svc_acc_change, incoming_updated_at).await, _ => Err(s3_error!( NotImplemented, "site replication IAM item type `{}` is not supported", @@ -9607,6 +9381,252 @@ async fn apply_iam_item(item: SRIAMItem) -> S3Result<()> { } } +async fn apply_iam_policy_item(iam_sys: &IamSys, name: &str, policy: Option) -> S3Result<()> { + if let Some(policy) = policy { + let policy: Policy = + serde_json::from_value(policy).map_err(|e| s3_error!(InvalidRequest, "invalid policy body: {}", e))?; + iam_sys.set_policy(name, policy).await.map_err(ApiError::from)?; + } else { + iam_sys.delete_policy(name, true).await.map_err(ApiError::from)?; + } + Ok(()) +} + +async fn apply_iam_policy_mapping_item(iam_sys: &IamSys, policy_mapping: Option) -> S3Result<()> { + let Some(mapping) = policy_mapping else { + return Err(s3_error!(InvalidRequest, "policyMapping is required")); + }; + let user_type = user_type_from_sr_wire(mapping.user_type).ok_or_else(|| s3_error!(InvalidRequest, "invalid userType"))?; + iam_sys + .policy_db_set(&mapping.user_or_group, user_type, mapping.is_group, &mapping.policy) + .await + .map_err(ApiError::from)?; + Ok(()) +} + +async fn apply_iam_group_info_item(iam_sys: &IamSys, group_info: Option) -> S3Result<()> { + let Some(group_info) = group_info else { + return Err(s3_error!(InvalidRequest, "groupInfo is required")); + }; + let update = group_info.update_req; + if !group_info_requires_upsert(&update) { + iam_sys + .remove_users_from_group(&update.group, update.members) + .await + .map_err(ApiError::from)?; + return Ok(()); + } + + iam_sys + .add_users_to_group(&update.group, update.members) + .await + .map_err(ApiError::from)?; + iam_sys + .set_group_status(&update.group, matches!(update.status, GroupStatus::Enabled)) + .await + .map_err(ApiError::from)?; + Ok(()) +} + +async fn apply_iam_sts_account_item(iam_sys: &IamSys, sts_credential: Option) -> S3Result<()> { + let Some(sts_credential) = sts_credential else { + return Err(s3_error!(InvalidRequest, "stsCredential is required")); + }; + let Some(secret) = current_token_signing_key() else { + return Err(s3_error!(InvalidRequest, "token signing key not initialized")); + }; + let claims = get_claims_from_token_with_secret(&sts_credential.session_token, &secret) + .map_err(|e| s3_error!(InvalidRequest, "invalid STS session token: {e}"))?; + let expiration = claims + .get("exp") + .and_then(claims_unix_timestamp) + .map(OffsetDateTime::from_unix_timestamp) + .transpose() + .map_err(|e| s3_error!(InvalidRequest, "invalid STS expiry: {e}"))?; + let groups = string_list_claim(&claims, "groups"); + let compatibility_policy = sts_replication_compatibility_policy(&claims, &sts_credential.parent_policy_mapping); + let cred = rustfs_credentials::Credentials { + access_key: sts_credential.access_key.clone(), + secret_key: sts_credential.secret_key.clone(), + session_token: sts_credential.session_token.clone(), + expiration, + status: "on".to_string(), + parent_user: sts_credential.parent_user.clone(), + groups, + claims: Some(claims), + ..Default::default() + }; + iam_sys + .set_temp_user(&sts_credential.access_key, &cred, compatibility_policy) + .await + .map_err(ApiError::from)?; + Ok(()) +} + +async fn apply_iam_user_item( + iam_sys: &IamSys, + iam_user: Option, + incoming_updated_at: Option, +) -> S3Result<()> { + let Some(user) = iam_user else { + return Err(s3_error!(InvalidRequest, "iamUser is required")); + }; + if let Some(local) = iam_sys.get_user(&user.access_key).await + && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) + { + return Ok(()); + } + if user.is_delete_req { + iam_sys.delete_user(&user.access_key, true).await.map_err(ApiError::from)?; + } else { + let Some(user_req) = user.user_req else { + return Err(s3_error!(InvalidRequest, "userReq is required")); + }; + let is_status_only_update = user_req.secret_key.is_empty() && user_req.policy.is_none(); + if is_status_only_update { + iam_sys + .set_user_status(&user.access_key, user_req.status) + .await + .map_err(ApiError::from)?; + } else { + iam_sys + .create_user(&user.access_key, &user_req) + .await + .map_err(ApiError::from)?; + } + } + Ok(()) +} + +async fn apply_iam_service_account_item( + iam_sys: &IamSys, + svc_acc_change: Option, + incoming_updated_at: Option, +) -> S3Result<()> { + let Some(change) = svc_acc_change else { + return Err(s3_error!(InvalidRequest, "serviceAccountChange is required")); + }; + let envelope = change.oidc_service_account_envelope; + if let Some(create) = change.create { + let local_updated_at = iam_sys + .get_user(&create.access_key) + .await + .map(|local| local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH)); + let replicated_policy = if create.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT { + if local_updated_at.is_some_and(|local_updated_at| is_stale_update(local_updated_at, incoming_updated_at)) { + return Ok(()); + } + ReplicatedServiceAccountPolicy { + policy: Some(site_replicator_service_account_policy()?), + is_envelope: false, + } + } else { + let Some(replicated_policy) = + decode_service_account_replication_policy(&create, envelope.as_ref(), incoming_updated_at, local_updated_at)? + else { + return Ok(()); + }; + replicated_policy + }; + match iam_sys.get_service_account(&create.access_key).await { + Ok((existing, _)) => { + if existing.parent_user != create.parent { + return Err(s3_error!( + InvalidRequest, + "service account {} already exists with a different parent user", + create.access_key + )); + } + iam_sys + .update_service_account( + &create.access_key, + UpdateServiceAccountOpts { + name: replicated_policy.metadata_for_existing_account(create.name), + description: replicated_policy.metadata_for_existing_account(create.description), + session_policy: replicated_policy.for_existing_account(), + secret_key: Some(create.secret_key), + expiration: create.expiration, + status: (!create.status.is_empty()).then_some(create.status), + parent_user: None, + allow_site_replicator_account: create.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT, + }, + ) + .await + .map_err(ApiError::from)?; + } + Err(err) if is_err_no_such_service_account(&err) => { + iam_sys + .new_service_account( + &create.parent, + Some(create.groups), + NewServiceAccountOpts { + session_policy: replicated_policy.policy, + access_key: create.access_key, + secret_key: create.secret_key, + name: (!create.name.is_empty()).then_some(create.name), + description: (!create.description.is_empty()).then_some(create.description), + expiration: create.expiration, + allow_site_replicator_account: true, + claims: Some(create.claims), + }, + ) + .await + .map_err(ApiError::from)?; + } + Err(err) => return Err(ApiError::from(err).into()), + } + return Ok(()); + } + + if let Some(update) = change.update { + if let Some(local) = iam_sys.get_user(&update.access_key).await + && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) + { + return Ok(()); + } + let allow_site_replicator_account = update.access_key == SITE_REPLICATOR_SERVICE_ACCOUNT; + let session_policy = if allow_site_replicator_account { + Some(site_replicator_service_account_policy()?) + } else { + update.session_policy.as_str().and_then(|raw| serde_json::from_str(raw).ok()) + }; + iam_sys + .update_service_account( + &update.access_key, + UpdateServiceAccountOpts { + session_policy, + secret_key: (!update.secret_key.is_empty()).then_some(update.secret_key), + name: (!update.name.is_empty()).then_some(update.name), + description: (!update.description.is_empty()).then_some(update.description), + expiration: update.expiration, + status: (!update.status.is_empty()).then_some(update.status), + // Peers replicate credentials, never the local parent binding: + // each site resolves its own parent from its own IAM. + parent_user: None, + allow_site_replicator_account, + }, + ) + .await + .map_err(ApiError::from)?; + return Ok(()); + } + + if let Some(delete) = change.delete { + if let Some(local) = iam_sys.get_user(&delete.access_key).await + && is_stale_update(local.update_at.unwrap_or(OffsetDateTime::UNIX_EPOCH), incoming_updated_at) + { + return Ok(()); + } + iam_sys + .delete_service_account(&delete.access_key, true) + .await + .map_err(ApiError::from)?; + return Ok(()); + } + + Err(s3_error!(InvalidRequest, "serviceAccountChange is empty")) +} + fn claims_unix_timestamp(value: &Value) -> Option { match value { Value::Number(number) => number.as_i64(), diff --git a/rustfs/src/admin/router.rs b/rustfs/src/admin/router.rs index 717de0bee..14ad1fb5e 100644 --- a/rustfs/src/admin/router.rs +++ b/rustfs/src/admin/router.rs @@ -17,7 +17,8 @@ use super::storage_api::bucket::metadata_sys; use super::storage_api::bucket::replication::{self, BucketReplicationResyncStatus, BucketStats, ReplicationStatusType}; use super::storage_api::bucket::target::{BucketTarget, BucketTargetType, BucketTargets}; use super::storage_api::bucket::target_sys::{ - BucketTargetSys, PutObjectOptions, RemoveObjectOptions, S3ClientError, TargetClient, append_version_id_query, + BucketTargetSys, PutObjectOptions, RemoveObjectOptions, S3ClientError, SsecPassthroughCapability, TargetClient, + append_version_id_query, }; use super::storage_api::bucket::versioning_sys::BucketVersioningSys; use super::storage_api::bucket::{AdminReplicationConfigExt as _, AdminVersioningConfigExt as _}; @@ -70,6 +71,9 @@ use rustfs_utils::egress::{OutboundDnsResolver, OutboundPolicy}; use rustfs_utils::http::headers::{ AMZ_CHECKSUM_CRC32, AMZ_CHECKSUM_CRC32C, AMZ_CHECKSUM_CRC64NVME, AMZ_CHECKSUM_SHA1, AMZ_CHECKSUM_SHA256, AMZ_CHECKSUM_TYPE, }; +use rustfs_utils::http::object_encryption_keys::{ + REPLICATION_SSEC_ALGORITHM_HEADER, REPLICATION_SSEC_KEY_MD5_HEADER, REPLICATION_SSEC_ORIGINAL_SIZE_HEADER, +}; use rustfs_utils::http::{ SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_REPLICATION_CHECK, SUFFIX_SOURCE_REPLICATION_REQUEST, SUFFIX_SOURCE_VERSION_ID, get_source_scheme, insert_header, @@ -213,6 +217,13 @@ const REPLICATION_CHECK_ERROR_MAX_BYTES: usize = 512; /// RustFS extension code (no madmin analogue): the target does not adopt the /// source version id, breaking the version-identity replication contract. const REPLICATION_CHECK_CODE_VERSION_MISMATCH: &str = "BucketRemoteTargetVersionMismatch"; +/// RustFS extension code (no madmin analogue): the target drops the +/// `X-Rustfs-Replication-*` SSE-C passthrough headers, so an SSE-C replica +/// would lose its decryption material (N2 fail-closed). +const REPLICATION_CHECK_CODE_SSEC_PASSTHROUGH: &str = "BucketRemoteSsecPassthroughUnsupported"; +/// Syntactically valid stand-in SSE-C key MD5 for the passthrough probe (the +/// probe object is never decrypted; it only has to round-trip the metadata). +const REPLICATION_CHECK_SSEC_PROBE_KEY_MD5: &str = "AAAAAAAAAAAAAAAAAAAAAA=="; #[derive(Debug, Clone, serde::Serialize)] struct ReplicationCheckResponse { @@ -254,6 +265,8 @@ struct ReplicationCheckPhases { put: ReplicationCheckPhaseStatus, #[serde(rename = "VersionFidelity")] version_fidelity: ReplicationCheckPhaseStatus, + #[serde(rename = "SsecPassthrough")] + ssec_passthrough: ReplicationCheckPhaseStatus, #[serde(rename = "DeleteMarker")] delete_marker: ReplicationCheckPhaseStatus, #[serde(rename = "VersionDelete")] @@ -1852,7 +1865,7 @@ fn build_replication_check_response(mut targets: Vec { + BucketTargetSys::get() + .record_ssec_passthrough_capability(&target.arn, SsecPassthroughCapability::Supported) + .await; + } + ("FAILED", Some(REPLICATION_CHECK_CODE_SSEC_PASSTHROUGH)) => { + BucketTargetSys::get() + .record_ssec_passthrough_capability(&target.arn, SsecPassthroughCapability::Unsupported) + .await; + } + _ => {} + } + result } @@ -2087,6 +2119,15 @@ struct ReplicationProbePutOutcome { response_version_id: Option, } +/// Outcome of the SSE-C passthrough probe: whether the HEAD-back of the probe +/// replica echoed SSE-C evidence (the customer-algorithm header a RustFS +/// target restores from the passthrough transport headers), plus the version +/// the target assigned so cleanup can address it. +struct ReplicationSsecProbeOutcome { + evidence_present: bool, + version_id: Option, +} + struct ReplicationProbeMultipartError { primary: S3ClientError, cleanup_error: Option, @@ -2109,9 +2150,14 @@ trait ReplicationProbeOperations { /// there: a target can adopt PutObject version ids and still mint its own /// for CreateMultipartUpload. async fn multipart_put(&mut self) -> Result; + /// PUT a probe version carrying the SSE-C passthrough transport headers, + /// HEAD it back through the replication-check channel, and report whether + /// the SSE-C evidence survived. Cleanup of the created version is the + /// caller's job (the outcome carries its version id). + async fn ssec_passthrough_probe(&mut self) -> Result; async fn create_delete_marker(&mut self, version_id: Option<&str>) -> Result, S3ClientError>; async fn delete_version(&mut self, version_id: Option<&str>) -> Result<(), S3ClientError>; - async fn cleanup(&mut self, known_version_ids: [Option<&str>; 3]) -> Result<(), String>; + async fn cleanup(&mut self, known_version_ids: [Option<&str>; 4]) -> Result<(), String>; } struct RemoteReplicationProbeOperations<'a> { @@ -2131,6 +2177,10 @@ impl ReplicationProbeOperations for RemoteReplicationProbeOperations<'_> { multipart_put_replication_probe_object(self.client, self.bucket, self.key, self.time).await } + async fn ssec_passthrough_probe(&mut self) -> Result { + ssec_passthrough_probe_object(self.client, self.bucket, self.key, self.time).await + } + async fn create_delete_marker(&mut self, version_id: Option<&str>) -> Result, S3ClientError> { delete_replication_probe_object( self.client, @@ -2154,7 +2204,7 @@ impl ReplicationProbeOperations for RemoteReplicationProbeOperations<'_> { .map(|_| ()) } - async fn cleanup(&mut self, known_version_ids: [Option<&str>; 3]) -> Result<(), String> { + async fn cleanup(&mut self, known_version_ids: [Option<&str>; 4]) -> Result<(), String> { cleanup_replication_probe(self.client, self.bucket, self.key, known_version_ids).await } } @@ -2175,6 +2225,7 @@ fn version_fidelity_error(api: &str, outcome: &ReplicationProbePutOutcome) -> Op async fn execute_replication_probe(result: &mut ReplicationCheckTargetStatus, operations: &mut impl ReplicationProbeOperations) { let mut probe_version_id = None; let mut multipart_probe_version_id = None; + let mut ssec_probe_version_id = None; let mut delete_marker_version_id = None; let mut cleanup_required = true; let mut multipart_cleanup_error = None; @@ -2230,6 +2281,38 @@ async fn execute_replication_probe(result: &mut ReplicationCheckTargetStatus, op } } + // N2: probe SSE-C passthrough with the same transport headers live + // replication sends. A target that drops them (MinIO, generic S3) stores + // the probe as a plain object and echoes no SSE-C evidence on the + // HEAD-back; SSE-C replicas there would silently lose their decryption + // material, so the target must be flagged with a machine-readable code. + // Deliberately unlike VersionFidelity, a failed SsecPassthrough phase + // does NOT fail the target overall: version-identity drift breaks the + // replication contract for every object, while dropped SSE-C passthrough + // headers only limit a capability — a plaintext-only deployment against a + // MinIO target is perfectly healthy and must not turn red. The phase's + // own FAILED + machine-readable Code remains for madmin consumers (and + // the verdict still reaches the runtime capability cache). + if result.phases.put.status == "OK" && result.phases.version_fidelity.status == "OK" { + match operations.ssec_passthrough_probe().await { + Ok(outcome) => { + ssec_probe_version_id = outcome.version_id; + if outcome.evidence_present { + result.phases.ssec_passthrough = ReplicationCheckPhaseStatus::passed(); + } else { + let error = "target drops SSE-C passthrough replication headers; \ + SSE-C replicas would lose their decryption material on this target"; + result.phases.ssec_passthrough = + ReplicationCheckPhaseStatus::failed_with_code(error, REPLICATION_CHECK_CODE_SSEC_PASSTHROUGH); + } + } + Err(err) => { + let error = format_replication_check_client_error(&err, ReplicationCheckFailureContext::ReplicateObject); + result.phases.ssec_passthrough = ReplicationCheckPhaseStatus::failed(&error); + } + } + } + if result.phases.put.status == "OK" && result.phases.version_fidelity.status == "OK" { match operations.create_delete_marker(probe_version_id.as_deref()).await { Ok(version_id) => { @@ -2258,6 +2341,7 @@ async fn execute_replication_probe(result: &mut ReplicationCheckTargetStatus, op .cleanup([ probe_version_id.as_deref(), multipart_probe_version_id.as_deref(), + ssec_probe_version_id.as_deref(), delete_marker_version_id.as_deref(), ]) .await @@ -2552,6 +2636,72 @@ async fn put_replication_probe_object( }) } +/// PUT a fresh probe version carrying the SSE-C passthrough transport headers +/// (the wire shape live SSE-C replication uses), then HEAD it back through the +/// worker channel (replication-check exemption + proxy suppression). A RustFS +/// target restores the transport headers into stored SSE-C metadata and its +/// HEAD echoes `x-amz-server-side-encryption-customer-algorithm`; a target +/// that dropped the headers echoes nothing. The probe body is never SSE-C +/// encrypted — only the metadata round-trip matters — and the version is +/// deleted by the shared probe cleanup. +async fn ssec_passthrough_probe_object( + target_client: &TargetClient, + target_bucket: &str, + probe_key: &str, + now: OffsetDateTime, +) -> Result { + let options = build_replication_probe_put_options(now); + let sent_version_id = options.internal.source_version_id.clone(); + let mut headers = build_replication_probe_headers(&options); + // These are full wire names (not x-rustfs/x-minio suffixes), so they must + // be inserted verbatim — `insert_header` would mangle them. + for (name, value) in [ + (REPLICATION_SSEC_ALGORITHM_HEADER, "AES256"), + (REPLICATION_SSEC_KEY_MD5_HEADER, REPLICATION_CHECK_SSEC_PROBE_KEY_MD5), + (REPLICATION_SSEC_ORIGINAL_SIZE_HEADER, "8"), + ] { + let name = name + .parse::() + .map_err(|err| S3ClientError::new(format!("invalid ssec probe header name: {err}")))?; + let value = + HeaderValue::from_str(value).map_err(|err| S3ClientError::new(format!("invalid ssec probe header value: {err}")))?; + headers.insert(name, value); + } + + let query_version_id = sent_version_id.clone(); + let response = target_client + .client + .put_object() + .bucket(target_bucket) + .key(probe_key) + .content_length(8) + .body(AwsByteStream::from_static(b"aaaaaaaa")) + .customize() + .map_request(move |mut req| { + for (key, value) in headers.clone() { + req.headers_mut().insert(key.expect("operation should succeed"), value); + } + let uri = append_version_id_query(req.uri(), &query_version_id); + req.set_uri(uri).map_err(std::io::Error::other)?; + Result::<_, std::io::Error>::Ok(req) + }) + .send() + .await + .map_err(S3ClientError::from)?; + let version_id = response.version_id().map(ToOwned::to_owned); + + let head_version = version_id.clone().or_else(|| Some(sent_version_id.clone())); + let head = target_client + .head_object(target_bucket, probe_key, head_version) + .await + .map_err(S3ClientError::from)?; + + Ok(ReplicationSsecProbeOutcome { + evidence_present: head.sse_customer_algorithm().is_some_and(|algorithm| !algorithm.is_empty()), + version_id, + }) +} + async fn delete_replication_probe_object( target_client: &TargetClient, target_bucket: &str, @@ -3722,6 +3872,12 @@ mod tests { /// Same, for the multipart leg: a target may mirror PutObject ids and /// still mint its own at CreateMultipartUpload. minted_multipart_version_id: Option<&'static str>, + /// Transport failure of the SSE-C passthrough probe itself. + ssec_probe_error: Option<&'static str>, + /// Models a MinIO-like target that drops the SSE-C passthrough + /// headers: the probe HEAD-back echoes no SSE-C evidence. The default + /// (false) models a RustFS target that preserves them. + ssec_evidence_missing: bool, delete_marker_error: Option<&'static str>, version_delete_error: Option<&'static str>, cleanup_error: Option<&'static str>, @@ -3759,6 +3915,17 @@ mod tests { }) } + async fn ssec_passthrough_probe(&mut self) -> Result { + self.calls.push("ssec-probe"); + match self.ssec_probe_error { + Some(code) => Err(scripted_probe_error(code)), + None => Ok(ReplicationSsecProbeOutcome { + evidence_present: !self.ssec_evidence_missing, + version_id: Some("ssec-version".to_string()), + }), + } + } + async fn create_delete_marker(&mut self, _version_id: Option<&str>) -> Result, S3ClientError> { self.calls.push("delete-marker"); match self.delete_marker_error { @@ -3775,7 +3942,7 @@ mod tests { } } - async fn cleanup(&mut self, known_version_ids: [Option<&str>; 3]) -> Result<(), String> { + async fn cleanup(&mut self, known_version_ids: [Option<&str>; 4]) -> Result<(), String> { self.calls.push("cleanup"); self.cleanup_ids = known_version_ids .into_iter() @@ -3810,8 +3977,9 @@ mod tests { assert_eq!(result.phases.version_fidelity.code, Some(REPLICATION_CHECK_CODE_VERSION_MISMATCH)); assert_eq!(result.phases.delete_marker.status, "SKIPPED"); assert_eq!(result.phases.version_delete.status, "SKIPPED"); + assert_eq!(result.phases.ssec_passthrough.status, "SKIPPED"); assert_eq!(result.phases.cleanup.status, "OK"); - assert_eq!(operations.cleanup_ids, [Some("target-minted-version".to_string()), None, None]); + assert_eq!(operations.cleanup_ids, [Some("target-minted-version".to_string()), None, None, None]); } #[tokio::test] @@ -3824,6 +3992,70 @@ mod tests { assert_eq!(result.status, "OK"); assert_eq!(result.phases.version_fidelity.status, "OK"); assert_eq!(result.phases.version_fidelity.code, None); + assert_eq!(result.phases.ssec_passthrough.status, "OK"); + assert_eq!(result.phases.ssec_passthrough.code, None); + } + + /// N2: a target that drops the SSE-C passthrough transport headers must + /// fail the SsecPassthrough phase with the machine-readable code while the + /// target overall stays OK — deliberately unlike VersionFidelity: this is + /// a capability limit, not a broken replication contract, and a + /// plaintext-only deployment against such a target must not turn red. The + /// other mutation phases keep running and the probe version is cleaned up. + #[tokio::test] + async fn replication_probe_flags_ssec_passthrough_dropping_target_without_failing_target() { + let mut result = replication_check_target("arn:a", "OK", None); + let mut operations = ScriptedReplicationProbe { + ssec_evidence_missing: true, + ..Default::default() + }; + + execute_replication_probe(&mut result, &mut operations).await; + + assert_eq!( + operations.calls, + [ + "put", + "multipart-put", + "ssec-probe", + "delete-marker", + "version-delete", + "cleanup" + ] + ); + assert_eq!(result.status, "OK", "a capability-only failure must not fail the target overall"); + assert_eq!(result.error, None); + assert_eq!(result.phases.ssec_passthrough.status, "FAILED"); + assert_eq!(result.phases.ssec_passthrough.code, Some(REPLICATION_CHECK_CODE_SSEC_PASSTHROUGH)); + assert_eq!( + operations.cleanup_ids, + [ + Some("object-version".to_string()), + Some("multipart-version".to_string()), + Some("ssec-version".to_string()), + Some("marker-version".to_string()) + ] + ); + } + + /// A transport failure of the SSE-C probe is not evidence of a dropping + /// target: the phase fails without the capability code (the runtime cache + /// stays Unknown and the worker keeps auditing), and the target overall + /// stays OK. + #[tokio::test] + async fn replication_probe_ssec_transport_failure_carries_no_capability_code() { + let mut result = replication_check_target("arn:a", "OK", None); + let mut operations = ScriptedReplicationProbe { + ssec_probe_error: Some("InternalError"), + ..Default::default() + }; + + execute_replication_probe(&mut result, &mut operations).await; + + assert_eq!(result.status, "OK"); + assert_eq!(result.phases.ssec_passthrough.status, "FAILED"); + assert_eq!(result.phases.ssec_passthrough.code, None); + assert_eq!(operations.cleanup_ids[2], None, "a failed ssec probe leaves no version to clean"); } #[tokio::test] @@ -3854,12 +4086,23 @@ mod tests { execute_replication_probe(&mut result, &mut operations).await; - assert_eq!(operations.calls, ["put", "multipart-put", "delete-marker", "version-delete", "cleanup"]); + assert_eq!( + operations.calls, + [ + "put", + "multipart-put", + "ssec-probe", + "delete-marker", + "version-delete", + "cleanup" + ] + ); assert_eq!( operations.cleanup_ids, [ Some("object-version".to_string()), Some("multipart-version".to_string()), + Some("ssec-version".to_string()), None ] ); @@ -3879,12 +4122,23 @@ mod tests { execute_replication_probe(&mut result, &mut operations).await; - assert_eq!(operations.calls, ["put", "multipart-put", "delete-marker", "version-delete", "cleanup"]); + assert_eq!( + operations.calls, + [ + "put", + "multipart-put", + "ssec-probe", + "delete-marker", + "version-delete", + "cleanup" + ] + ); assert_eq!( operations.cleanup_ids, [ Some("object-version".to_string()), Some("multipart-version".to_string()), + Some("ssec-version".to_string()), Some("marker-version".to_string()) ] ); diff --git a/rustfs/src/admin/storage_api.rs b/rustfs/src/admin/storage_api.rs index dbf3a81df..eb38e9640 100644 --- a/rustfs/src/admin/storage_api.rs +++ b/rustfs/src/admin/storage_api.rs @@ -196,6 +196,7 @@ pub(crate) mod bucket_target_sys { pub(crate) type PutObjectOptions = super::ecstore_bucket::bucket_target_sys::PutObjectOptions; pub(crate) type RemoveObjectOptions = super::ecstore_bucket::bucket_target_sys::RemoveObjectOptions; pub(crate) type S3ClientError = super::ecstore_bucket::bucket_target_sys::S3ClientError; + pub(crate) type SsecPassthroughCapability = super::ecstore_bucket::bucket_target_sys::SsecPassthroughCapability; pub(crate) type TargetClient = super::ecstore_bucket::bucket_target_sys::TargetClient; } diff --git a/rustfs/src/app/object_usecase.rs b/rustfs/src/app/object_usecase.rs index bcce2b705..413a88464 100644 --- a/rustfs/src/app/object_usecase.rs +++ b/rustfs/src/app/object_usecase.rs @@ -46,9 +46,10 @@ use super::storage_api::object_usecase::bucket::{ replication::{ DeleteReplicationConfigSnapshot, REPLICATE_INCOMING_DELETE, ReplicationStatusType, commit_force_delete_intent, delete_replication_state_from_config, delete_replication_version_id, deleted_object_has_pending_replication_delete, - force_delete_target_set, has_active_delete_rule, load_delete_config_snapshot, must_replicate_object, - persist_force_delete_intent, schedule_object_replication, schedule_replication_delete, schedule_replication_deletes, - set_deleted_object_replication_state, should_schedule_delete_replication, should_use_existing_delete_replication_info, + force_delete_target_set, get_read_proxy_targets, has_active_delete_rule, load_delete_config_snapshot, + must_replicate_object, persist_force_delete_intent, record_replication_proxy, schedule_object_replication, + schedule_replication_delete, schedule_replication_deletes, set_deleted_object_replication_state, + should_schedule_delete_replication, should_use_existing_delete_replication_info, }, tagging::decode_tags, validate_restore_request, @@ -6662,6 +6663,226 @@ impl DefaultObjectUsecase { }) } + /// Headers a proxied read forwards verbatim to the replication target: + /// only the client's SSE-C key family, so the target performs the real + /// SSE-C decryption (never the replication-check exemption). HTTP + /// conditional headers (If-Match & co.) are deliberately NOT forwarded — + /// MinIO does not forward them either, and a remote 304/412 would leak a + /// conditional evaluation against a replica the local site never saw. + /// Range and part-number travel as typed SDK parameters instead. + fn proxy_read_passthrough_headers(headers: &HeaderMap) -> HeaderMap { + const FORWARDED: &[&str] = &[ + "x-amz-server-side-encryption-customer-algorithm", + "x-amz-server-side-encryption-customer-key", + "x-amz-server-side-encryption-customer-key-md5", + ]; + let mut forwarded = HeaderMap::new(); + for name in FORWARDED { + if let Ok(header_name) = http::HeaderName::from_str(name) + && let Some(value) = headers.get(&header_name) + { + forwarded.insert(header_name, value.clone()); + } + } + forwarded + } + + /// True when a proxied SDK call failed because the target does not have + /// the object either (service-level not-found or a raw 404, which also + /// covers NoSuchVersion): the caller tries the next target silently. + fn proxy_sdk_error_is_not_found(err: &aws_sdk_s3::error::SdkError) -> bool { + err.raw_response().is_some_and(|resp| resp.status().as_u16() == 404) + } + + /// Serve a GET whose local read failed with not-found by proxying to the + /// bucket's replication targets (MinIO `proxyGetToReplicationTarget`, + /// backlog#1675 P1-5). Returns None when no target can serve the object; + /// the caller then returns the original local error. + async fn proxy_get_object_to_replication_targets( + req: &S3Request, + bucket: &str, + key: &str, + opts: &ObjectOptions, + ) -> Option { + let targets = get_read_proxy_targets(bucket, key, opts).await; + if targets.is_empty() { + return None; + } + let extra_headers = Self::proxy_read_passthrough_headers(&req.headers); + let range = req + .headers + .get(http::header::RANGE) + .and_then(|value| value.to_str().ok()) + .map(str::to_owned); + let part_number = req.input.part_number; + + for target in targets { + match target + .get_object( + &target.bucket, + key, + opts.version_id.clone(), + range.clone(), + part_number, + extra_headers.clone(), + ) + .await + { + Ok(remote) => { + // MinIO-aligned accounting: one total per proxy attempt + // (targets were available), one failed when no target + // served it — never per target. + record_replication_proxy(bucket, "GetObject", false).await; + return Some(Self::proxy_sdk_get_output_to_s3s(remote)); + } + Err(err) if Self::proxy_sdk_error_is_not_found(&err) => { + debug!(bucket, key, arn = %target.arn, "read proxy: target does not have the object"); + } + Err(err) => { + warn!(bucket, key, arn = %target.arn, error = %err, "read proxy: GET against replication target failed"); + } + } + } + record_replication_proxy(bucket, "GetObject", true).await; + None + } + + /// Serve a HEAD whose local lookup failed with not-found by proxying to + /// the bucket's replication targets (MinIO `proxyHeadToRepTarget`). + async fn proxy_head_object_to_replication_targets( + req: &S3Request, + bucket: &str, + key: &str, + opts: &ObjectOptions, + ) -> Option { + let targets = get_read_proxy_targets(bucket, key, opts).await; + if targets.is_empty() { + return None; + } + let extra_headers = Self::proxy_read_passthrough_headers(&req.headers); + let range = req + .headers + .get(http::header::RANGE) + .and_then(|value| value.to_str().ok()) + .map(str::to_owned); + let part_number = req.input.part_number; + + for target in targets { + match target + .head_object_for_proxy( + &target.bucket, + key, + opts.version_id.clone(), + range.clone(), + part_number, + extra_headers.clone(), + ) + .await + { + Ok(remote) => { + // MinIO-aligned accounting: one total per proxy attempt, + // one failed when no target served it. + record_replication_proxy(bucket, "HeadObject", false).await; + return Some(Self::proxy_sdk_head_output_to_s3s(remote)); + } + Err(err) if Self::proxy_sdk_error_is_not_found(&err) => { + debug!(bucket, key, arn = %target.arn, "read proxy: target does not have the object"); + } + Err(err) => { + warn!(bucket, key, arn = %target.arn, error = %err, "read proxy: HEAD against replication target failed"); + } + } + } + record_replication_proxy(bucket, "HeadObject", true).await; + None + } + + /// Translate a proxied SDK GET response into the s3s output, forwarding + /// the body as a stream (no buffering, no local persistence). + fn proxy_sdk_get_output_to_s3s(remote: aws_sdk_s3::operation::get_object::GetObjectOutput) -> GetObjectOutput { + let body = remote.body; + let body_stream = tokio_util::io::ReaderStream::with_capacity(body.into_async_read(), 64 * 1024); + GetObjectOutput { + body: Some(StreamingBlob::wrap(body_stream)), + content_length: remote.content_length, + content_range: remote.content_range, + content_type: remote.content_type.as_deref().and_then(|v| ContentType::from_str(v).ok()), + content_encoding: remote.content_encoding, + content_disposition: remote.content_disposition, + content_language: remote.content_language, + cache_control: remote.cache_control, + accept_ranges: Some(ACCEPT_RANGES_BYTES.to_string()), + e_tag: remote.e_tag.as_deref().and_then(|v| ETag::from_str(v).ok()), + last_modified: remote + .last_modified + .and_then(|dt| OffsetDateTime::from_unix_timestamp_nanos(dt.as_nanos()).ok()) + .map(Timestamp::from), + metadata: remote.metadata, + version_id: remote.version_id, + server_side_encryption: remote + .server_side_encryption + .map(|sse| ServerSideEncryption::from(sse.as_str().to_string())), + sse_customer_algorithm: remote.sse_customer_algorithm, + sse_customer_key_md5: remote.sse_customer_key_md5, + ssekms_key_id: remote.ssekms_key_id, + parts_count: remote.parts_count, + tag_count: remote.tag_count, + storage_class: remote.storage_class.map(|sc| StorageClass::from(sc.as_str().to_string())), + expiration: remote.expiration, + restore: remote.restore, + checksum_crc32: remote.checksum_crc32, + checksum_crc32c: remote.checksum_crc32_c, + checksum_crc64nvme: remote.checksum_crc64_nvme, + checksum_sha1: remote.checksum_sha1, + checksum_sha256: remote.checksum_sha256, + checksum_type: remote.checksum_type.map(|ct| ChecksumType::from(ct.as_str().to_string())), + ..Default::default() + } + } + + /// Translate a proxied SDK HEAD response into the s3s output. + /// + /// Known gaps: the SDK's HeadObjectOutput does not model 206/Content-Range + /// for a ranged HEAD (the SDK exposes no content_range member on HEAD), + /// and s3s' typed HeadObjectOutput has no tag_count field (the local path + /// injects x-amz-tagging-count as a raw header) — both are dropped for + /// proxied HEADs. + fn proxy_sdk_head_output_to_s3s(remote: aws_sdk_s3::operation::head_object::HeadObjectOutput) -> HeadObjectOutput { + HeadObjectOutput { + content_length: remote.content_length, + content_type: remote.content_type.as_deref().and_then(|v| ContentType::from_str(v).ok()), + content_encoding: remote.content_encoding, + content_disposition: remote.content_disposition, + content_language: remote.content_language, + cache_control: remote.cache_control, + accept_ranges: Some(ACCEPT_RANGES_BYTES.to_string()), + e_tag: remote.e_tag.as_deref().and_then(|v| ETag::from_str(v).ok()), + last_modified: remote + .last_modified + .and_then(|dt| OffsetDateTime::from_unix_timestamp_nanos(dt.as_nanos()).ok()) + .map(Timestamp::from), + metadata: remote.metadata, + version_id: remote.version_id, + server_side_encryption: remote + .server_side_encryption + .map(|sse| ServerSideEncryption::from(sse.as_str().to_string())), + sse_customer_algorithm: remote.sse_customer_algorithm, + sse_customer_key_md5: remote.sse_customer_key_md5, + ssekms_key_id: remote.ssekms_key_id, + parts_count: remote.parts_count, + storage_class: remote.storage_class.map(|sc| StorageClass::from(sc.as_str().to_string())), + expiration: remote.expiration, + restore: remote.restore, + checksum_crc32: remote.checksum_crc32, + checksum_crc32c: remote.checksum_crc32_c, + checksum_crc64nvme: remote.checksum_crc64_nvme, + checksum_sha1: remote.checksum_sha1, + checksum_sha256: remote.checksum_sha256, + checksum_type: remote.checksum_type.map(|ct| ChecksumType::from(ct.as_str().to_string())), + ..Default::default() + } + } + #[instrument(name = "execute_get_object", level = "trace", skip(self, req))] pub async fn execute_get_object(&self, req: S3Request) -> S3Result> { self.execute_get_object_boxed(req).await @@ -6787,6 +7008,19 @@ impl DefaultObjectUsecase { { Ok(prepared_read) => prepared_read, Err(err) => { + // Active-active replication lag window: an object missing + // locally (and only missing — other errors keep their + // semantics) may still be served by proxying the GET to a + // replication target (backlog#1675 P1-5). + if matches!(*err.code(), S3ErrorCode::NoSuchKey | S3ErrorCode::NoSuchVersion) + && let Some(output) = Self::proxy_get_object_to_replication_targets(&req, &bucket, &key, &opts).await + { + lifecycle.finish_ok(); + let response = wrap_response_with_cors(&bucket, &req.method, &req.headers, output).await; + let result = Ok(response); + let _ = helper.version_id(version_id_for_event).complete(&result); + return result; + } lifecycle.finish_err(); return Err(err); } @@ -8696,6 +8930,17 @@ impl DefaultObjectUsecase { let msg = head_prefix_not_found_message(&bucket, &key, has_children); return Err(S3Error::with_message(S3ErrorCode::NoSuchKey, msg)); } + // Active-active replication lag window: an object missing + // locally may still be served by proxying the HEAD to a + // replication target (backlog#1675 P1-5). + if let Some(output) = Self::proxy_head_object_to_replication_targets(&req, &bucket, &key, &opts).await { + let response = wrap_response_with_cors(&bucket, &req.method, &req.headers, output).await; + let result = Ok(response); + let _ = helper + .version_id(req.input.version_id.clone().unwrap_or_default()) + .complete(&result); + return result; + } return Err(S3Error::new(S3ErrorCode::NoSuchKey)); } // Other errors, such as insufficient permissions, still return the original error diff --git a/rustfs/src/app/storage_api.rs b/rustfs/src/app/storage_api.rs index c6f03f240..0c2b3d4d4 100644 --- a/rustfs/src/app/storage_api.rs +++ b/rustfs/src/app/storage_api.rs @@ -627,6 +627,24 @@ pub(crate) mod bucket { #[cfg(test)] pub(crate) use replication_contracts::replication_statuses_map; + /// Remote replication-target client used by the read-proxy path. + pub(crate) type ProxyTargetClient = crate::storage::storage_api::ecstore_bucket::bucket_target_sys::TargetClient; + + /// Proxy-request metric recorder (get/head/tagging totals + failures). + pub(crate) use crate::storage::storage_api::record_replication_proxy; + + /// Replication targets eligible to serve a proxied GET/HEAD/Tagging of + /// an object not present locally (MinIO `getProxyTargets`; empty when + /// the request was itself proxied, versioning is suspended, or no + /// replication rule matches). backlog#1675 P1-5. + pub(crate) async fn get_read_proxy_targets( + bucket: &str, + object: &str, + opts: &crate::storage::storage_api::StorageObjectOptions, + ) -> Vec> { + replication_contracts::get_proxy_targets(bucket, object, opts).await + } + pub(crate) async fn persist_force_delete_intent( store: Arc, bucket: String, diff --git a/rustfs/src/storage/ecfs.rs b/rustfs/src/storage/ecfs.rs index 9369d7224..da0c9b775 100644 --- a/rustfs/src/storage/ecfs.rs +++ b/rustfs/src/storage/ecfs.rs @@ -12,15 +12,15 @@ // See the License for the specific language governing permissions and // limitations under the License. +use super::StorageVersioningConfigExt as _; use super::{ BUCKET_ACCELERATE_CONFIG, BUCKET_LOGGING_CONFIG, BUCKET_REQUEST_PAYMENT_CONFIG, BUCKET_VERSIONING_CONFIG, BUCKET_WEBSITE_CONFIG, BucketVersioningSys, OBJECT_LOCK_CONFIG, StorageError, check_retention_for_modification, decode_tags, decode_tags_to_map, delete_bucket_metadata_config_if_incarnation, encode_tags, get_bucket_accelerate_config, - get_bucket_logging_config, get_bucket_object_lock_config, get_bucket_replication_config, get_bucket_request_payment_config, - get_bucket_website_config, is_err_bucket_not_found, is_err_object_not_found, is_err_version_not_found, - record_replication_proxy, serialize, update_bucket_metadata_config_if_incarnation, + get_bucket_logging_config, get_bucket_object_lock_config, get_bucket_request_payment_config, get_bucket_website_config, + is_err_bucket_not_found, is_err_object_not_found, is_err_version_not_found, record_replication_proxy, serialize, + update_bucket_metadata_config_if_incarnation, }; -use super::{StorageReplicationConfigExt as _, StorageVersioningConfigExt as _}; use crate::admin::handlers::site_replication::site_replication_bucket_meta_hook; use crate::error::ApiError; use crate::storage::access::{apply_bucket_generation_guard, bucket_config_mutation_incarnation, has_bypass_governance_header}; @@ -59,7 +59,7 @@ const LOG_SUBSYSTEM_OBJECT_LOCK: &str = "object_lock"; const LOG_SUBSYSTEM_TAGGING: &str = "tagging"; use crate::app::storage_api::object_usecase::bucket::replication::{ - ReplicateDecision, must_replicate_metadata, schedule_metadata_replication, + ReplicateDecision, get_read_proxy_targets, must_replicate_metadata, schedule_metadata_replication, }; use crate::storage::storage_api::ecfs_consumer::StorageObjectOptions as ObjectOptions; @@ -105,18 +105,152 @@ impl FS { &self.server_ctx } - async fn replication_tagging_enabled(bucket: &str, object: &str) -> bool { - get_bucket_replication_config(bucket) - .await - .map(|(cfg, _)| cfg.has_active_rules(object, true)) - .unwrap_or(false) + /// Not-found classifier for proxied SDK tagging calls: a raw 404 covers + /// NoSuchKey and NoSuchVersion alike; the caller silently tries the next + /// replication target. + fn proxy_sdk_error_is_not_found(err: &aws_sdk_s3::error::SdkError) -> bool { + err.raw_response().is_some_and(|resp| resp.status().as_u16() == 404) } - async fn record_replication_tagging_metric(bucket: &str, object: &str, api: &str, is_err: bool) { - if !Self::replication_tagging_enabled(bucket, object).await { - return; + /// Selector options for a tagging proxy. Reuses `get_opts` so the + /// anti-loop `source-proxy-request` header family and the bucket's + /// version-suspension state gate proxying exactly like GET/HEAD. + async fn tagging_proxy_opts( + bucket: &str, + object: &str, + version_id: Option, + headers: &http::HeaderMap, + ) -> Option { + get_opts(bucket, object, version_id, None, headers).await.ok() + } + + /// Serve a GetObjectTagging for an object missing locally by proxying to + /// the bucket's replication targets (MinIO `proxyGetTaggingToRepTarget`, + /// backlog#1675 P1-5). None means no target had the object. + async fn proxy_get_object_tagging( + bucket: &str, + object: &str, + version_id: Option, + headers: &http::HeaderMap, + ) -> Option { + let opts = Self::tagging_proxy_opts(bucket, object, version_id, headers).await?; + let targets = get_read_proxy_targets(bucket, object, &opts).await; + if targets.is_empty() { + return None; } - record_replication_proxy(bucket, api, is_err).await; + for target in targets { + match target + .get_object_tagging(&target.bucket, object, opts.version_id.clone()) + .await + { + Ok(remote) => { + // MinIO-aligned accounting: one total per proxy attempt, + // one failed when no target served it. + record_replication_proxy(bucket, "GetObjectTagging", false).await; + return Some( + remote + .tag_set + .into_iter() + .map(|tag| Tag { + key: Some(tag.key), + value: Some(tag.value), + }) + .collect(), + ); + } + Err(err) if Self::proxy_sdk_error_is_not_found(&err) => { + debug!(bucket, object, arn = %target.arn, "tagging proxy: target does not have the object"); + } + Err(err) => { + warn!(bucket, object, arn = %target.arn, error = %err, "tagging proxy: GetObjectTagging against replication target failed"); + } + } + } + record_replication_proxy(bucket, "GetObjectTagging", true).await; + None + } + + /// Apply a PutObjectTagging for an object missing locally on a + /// replication target (MinIO `proxyTaggingToRepTarget`). + async fn proxy_put_object_tagging( + bucket: &str, + object: &str, + version_id: Option, + headers: &http::HeaderMap, + tag_set: &TagSet, + ) -> Option<()> { + let opts = Self::tagging_proxy_opts(bucket, object, version_id, headers).await?; + let mut tagging = aws_sdk_s3::types::Tagging::builder(); + for tag in tag_set { + let sdk_tag = aws_sdk_s3::types::Tag::builder() + .key(tag.key.clone().unwrap_or_default()) + .value(tag.value.clone().unwrap_or_default()) + .build() + .ok()?; + tagging = tagging.tag_set(sdk_tag); + } + let tagging = tagging.build().ok()?; + let targets = get_read_proxy_targets(bucket, object, &opts).await; + if targets.is_empty() { + return None; + } + for target in targets { + match target + .put_object_tagging(&target.bucket, object, opts.version_id.clone(), tagging.clone()) + .await + { + Ok(_) => { + // MinIO-aligned accounting: one total per proxy attempt, + // one failed when no target served it. + record_replication_proxy(bucket, "PutObjectTagging", false).await; + return Some(()); + } + Err(err) if Self::proxy_sdk_error_is_not_found(&err) => { + debug!(bucket, object, arn = %target.arn, "tagging proxy: target does not have the object"); + } + Err(err) => { + warn!(bucket, object, arn = %target.arn, error = %err, "tagging proxy: PutObjectTagging against replication target failed"); + } + } + } + record_replication_proxy(bucket, "PutObjectTagging", true).await; + None + } + + /// Apply a DeleteObjectTagging for an object missing locally on a + /// replication target (MinIO `proxyTaggingToRepTarget`). + async fn proxy_delete_object_tagging( + bucket: &str, + object: &str, + version_id: Option, + headers: &http::HeaderMap, + ) -> Option<()> { + let opts = Self::tagging_proxy_opts(bucket, object, version_id, headers).await?; + let targets = get_read_proxy_targets(bucket, object, &opts).await; + if targets.is_empty() { + return None; + } + for target in targets { + match target + .delete_object_tagging(&target.bucket, object, opts.version_id.clone()) + .await + { + Ok(_) => { + // MinIO-aligned accounting: one total per proxy attempt, + // one failed when no target served it. + record_replication_proxy(bucket, "DeleteObjectTagging", false).await; + return Some(()); + } + Err(err) if Self::proxy_sdk_error_is_not_found(&err) => { + debug!(bucket, object, arn = %target.arn, "tagging proxy: target does not have the object"); + } + Err(err) => { + warn!(bucket, object, arn = %target.arn, error = %err, "tagging proxy: DeleteObjectTagging against replication target failed"); + } + } + } + record_replication_proxy(bucket, "DeleteObjectTagging", true).await; + None } pub async fn get_object_tag_conditions_for_policy( @@ -447,7 +581,27 @@ impl S3 for FS { let mut opts = get_opts(&bucket, &object, version_id.clone(), None, &req.headers) .await .map_err(ApiError::from)?; - let existing_object_info = store.get_object_info(&bucket, &object, &opts).await.map_err(ApiError::from)?; + let existing_object_info = match store.get_object_info(&bucket, &object, &opts).await { + Ok(info) => info, + Err(e) => { + // Replication lag window: apply the tagging delete on a + // replication target that already has the object + // (backlog#1675 P1-5). No local object exists, so no bucket + // notification event is emitted for the proxied write. + if (is_err_object_not_found(&e) || is_err_version_not_found(&e)) + && Self::proxy_delete_object_tagging(&bucket, &object, version_id.clone(), &req.headers) + .await + .is_some() + { + counter!("rustfs_delete_object_tagging_success").increment(1); + let duration = start_time.elapsed(); + histogram!("rustfs_object_tagging_operation_duration_seconds", "operation" => "delete") + .record(duration.as_secs_f64()); + return Ok(S3Response::new(DeleteObjectTaggingOutput { version_id })); + } + return Err(ApiError::from(e).into()); + } + }; let dsc = must_replicate_metadata( &bucket, &object, @@ -470,7 +624,6 @@ impl S3 for FS { } let delete_tags_result = store.delete_object_tags(&bucket, &object, &opts).await; - Self::record_replication_tagging_metric(&bucket, &object, "DeleteObjectTagging", delete_tags_result.is_err()).await; let object_info = delete_tags_result.map_err(|e| { error!( component = LOG_COMPONENT_STORAGE, @@ -928,32 +1081,49 @@ impl S3 for FS { ..Default::default() }; - let tags_result = store.get_object_tags(bucket, object, &opts).await; - Self::record_replication_tagging_metric(bucket, object, "GetObjectTagging", tags_result.is_err()).await; - let tags = tags_result.map_err(|e| { - if is_err_object_not_found(&e) { - debug!( + let tags = match store.get_object_tags(bucket, object, &opts).await { + Ok(tags) => tags, + Err(e) => { + // Replication lag window: the object may exist on a + // replication target even though it is missing locally — + // proxy the tagging read there (backlog#1675 P1-5). + if (is_err_object_not_found(&e) || is_err_version_not_found(&e)) + && let Some(tag_set) = + Self::proxy_get_object_tagging(bucket, object, req.input.version_id.clone(), &req.headers).await + { + counter!("rustfs_get_object_tagging_success").increment(1); + let duration = start_time.elapsed(); + histogram!("rustfs_object_tagging_operation_duration_seconds", "operation" => "get") + .record(duration.as_secs_f64()); + return Ok(S3Response::new(GetObjectTaggingOutput { + tag_set, + version_id: req.input.version_id.clone(), + })); + } + if is_err_object_not_found(&e) { + debug!( + component = LOG_COMPONENT_STORAGE, + subsystem = LOG_SUBSYSTEM_TAGGING, + event = "object_tagging_not_found", + bucket = %bucket, + object = %object, + error = %e, + "Object tags not found" + ); + return Err(s3_error!(NoSuchKey)); + } + error!( component = LOG_COMPONENT_STORAGE, subsystem = LOG_SUBSYSTEM_TAGGING, - event = "object_tagging_not_found", + event = "object_tagging_get_failed", bucket = %bucket, object = %object, error = %e, - "Object tags not found" + "Failed to load object tags" ); - return s3_error!(NoSuchKey); + return Err(ApiError::from(e).into()); } - error!( - component = LOG_COMPONENT_STORAGE, - subsystem = LOG_SUBSYSTEM_TAGGING, - event = "object_tagging_get_failed", - bucket = %bucket, - object = %object, - error = %e, - "Failed to load object tags" - ); - ApiError::from(e).into() - })?; + }; let tag_set = decode_tags(tags.as_str()); debug!( @@ -1629,14 +1799,36 @@ impl S3 for FS { return Err(S3Error::with_message(S3ErrorCode::InternalError, "Not init".to_string())); }; - let tags = encode_tags(tagging.tag_set); + let tags = encode_tags(tagging.tag_set.clone()); debug!("Encoded tags: {}", tags); let version_id = req.input.version_id.clone(); let mut opts = get_opts(&bucket, &object, version_id.clone(), None, &req.headers) .await .map_err(ApiError::from)?; - let existing_object_info = store.get_object_info(&bucket, &object, &opts).await.map_err(ApiError::from)?; + let existing_object_info = match store.get_object_info(&bucket, &object, &opts).await { + Ok(info) => info, + Err(e) => { + // Replication lag window: apply the tagging update on a + // replication target that already has the object + // (backlog#1675 P1-5). No local object exists, so no bucket + // notification event is emitted for the proxied write. + if (is_err_object_not_found(&e) || is_err_version_not_found(&e)) + && Self::proxy_put_object_tagging(&bucket, &object, version_id.clone(), &req.headers, &tagging.tag_set) + .await + .is_some() + { + counter!("rustfs_put_object_tagging_success").increment(1); + let duration = start_time.elapsed(); + histogram!("rustfs_object_tagging_operation_duration_seconds", "operation" => "put") + .record(duration.as_secs_f64()); + return Ok(S3Response::new(PutObjectTaggingOutput { + version_id: req.input.version_id.clone(), + })); + } + return Err(ApiError::from(e).into()); + } + }; let dsc = must_replicate_metadata( &bucket, &object, @@ -1659,7 +1851,6 @@ impl S3 for FS { } let put_tags_result = store.put_object_tags(&bucket, &object, &tags, &opts).await; - Self::record_replication_tagging_metric(&bucket, &object, "PutObjectTagging", put_tags_result.is_err()).await; let object_info = put_tags_result.map_err(|e| { error!("Failed to put object tags: {}", e); counter!("rustfs_put_object_tagging_failure").increment(1); diff --git a/rustfs/src/storage/mod.rs b/rustfs/src/storage/mod.rs index e0817da90..f1fe52979 100644 --- a/rustfs/src/storage/mod.rs +++ b/rustfs/src/storage/mod.rs @@ -55,24 +55,24 @@ pub(crate) use storage_api::{ QuotaError, RUSTFS_META_BUCKET, RawFileInfo, ReadMultipleReq, ReadMultipleResp, ReadOptions, RenameDataResp, ReplicationStats, ReplicationStatusType, Result, SERVICE_SIGNAL_REFRESH_CONFIG, SERVICE_SIGNAL_RELOAD_DYNAMIC, StorageDeletedObject, StorageDiskRpcExt, StorageError, StorageGetObjectReader, StorageObjectInfo, StorageObjectOptions, - StorageObjectToDelete, StoragePeerS3ClientExt, StoragePutObjReader, StorageReplicationConfigExt, StorageVersioningConfigExt, - TONIC_RPC_PREFIX, TierConfigMgr, UpdateMetadataOpts, VolumeInfo, WalkDirOptions, WorkloadAdmissionSnapshotProviderRef, - WriteEncryption, WritePlan, access_consumer, add_object_lock_years, all_local_disk, all_local_disk_path, - check_retention_for_modification, collect_local_metrics, compression_metadata_value, contract, decode_tags, - decode_tags_to_map, delete_bucket_metadata_config, delete_bucket_metadata_config_if_incarnation, disk_drive_path, - disk_endpoint, ecfs_consumer, ecfs_extend_consumer, ecstore_admin, ecstore_bucket, ecstore_capacity, ecstore_client, - ecstore_cluster, ecstore_compression, ecstore_config, ecstore_data_usage, ecstore_disk, ecstore_error, ecstore_event, - ecstore_layout, ecstore_metrics, ecstore_notification, ecstore_rebalance, ecstore_rio, ecstore_rpc, ecstore_set_disk, - ecstore_storage, ecstore_tier, encode_tags, find_local_disk_by_ref, get_bucket_accelerate_config, get_bucket_cors_config, - get_bucket_logging_config, get_bucket_metadata, get_bucket_notification_config, get_bucket_object_lock_config, - get_bucket_replication_config, get_bucket_request_payment_config, get_bucket_sse_config, get_bucket_website_config, - get_local_server_property, get_lock_acquire_timeout, head_prefix_consumer, helper_consumer, init_background_replication, - init_bucket_metadata_sys, init_ecstore_config, init_local_disks_with_instance_ctx, init_lock_clients, - is_err_bucket_not_found, is_err_object_not_found, is_err_version_not_found, is_valid_storage_class, options_consumer, - prewarm_local_disk_id_map_with_instance_ctx, read_config, record_replication_proxy, rpc_consumer, runtime_sources_consumer, - s3_api_consumer, serialize, table_catalog_path_hash, to_s3s_etag, topology_snapshot_from_endpoint_pools_with_capabilities, - try_migrate_bucket_metadata, try_migrate_iam_config, try_migrate_server_config, update_bucket_metadata_config, - update_bucket_metadata_config_if_incarnation, verify_rpc_signature, wrap_reader, + StorageObjectToDelete, StoragePeerS3ClientExt, StoragePutObjReader, StorageVersioningConfigExt, TONIC_RPC_PREFIX, + TierConfigMgr, UpdateMetadataOpts, VolumeInfo, WalkDirOptions, WorkloadAdmissionSnapshotProviderRef, WriteEncryption, + WritePlan, access_consumer, add_object_lock_years, all_local_disk, all_local_disk_path, check_retention_for_modification, + collect_local_metrics, compression_metadata_value, contract, decode_tags, decode_tags_to_map, delete_bucket_metadata_config, + delete_bucket_metadata_config_if_incarnation, disk_drive_path, disk_endpoint, ecfs_consumer, ecfs_extend_consumer, + ecstore_admin, ecstore_bucket, ecstore_capacity, ecstore_client, ecstore_cluster, ecstore_compression, ecstore_config, + ecstore_data_usage, ecstore_disk, ecstore_error, ecstore_event, ecstore_layout, ecstore_metrics, ecstore_notification, + ecstore_rebalance, ecstore_rio, ecstore_rpc, ecstore_set_disk, ecstore_storage, ecstore_tier, encode_tags, + find_local_disk_by_ref, get_bucket_accelerate_config, get_bucket_cors_config, get_bucket_logging_config, get_bucket_metadata, + get_bucket_notification_config, get_bucket_object_lock_config, get_bucket_request_payment_config, get_bucket_sse_config, + get_bucket_website_config, get_local_server_property, get_lock_acquire_timeout, head_prefix_consumer, helper_consumer, + init_background_replication, init_bucket_metadata_sys, init_ecstore_config, init_local_disks_with_instance_ctx, + init_lock_clients, is_err_bucket_not_found, is_err_object_not_found, is_err_version_not_found, is_valid_storage_class, + options_consumer, prewarm_local_disk_id_map_with_instance_ctx, read_config, record_replication_proxy, rpc_consumer, + runtime_sources_consumer, s3_api_consumer, serialize, table_catalog_path_hash, to_s3s_etag, + topology_snapshot_from_endpoint_pools_with_capabilities, try_migrate_bucket_metadata, try_migrate_iam_config, + try_migrate_server_config, update_bucket_metadata_config, update_bucket_metadata_config_if_incarnation, verify_rpc_signature, + wrap_reader, }; #[cfg(test)] diff --git a/rustfs/src/storage/options.rs b/rustfs/src/storage/options.rs index ac920b5c3..dc5a6ca19 100644 --- a/rustfs/src/storage/options.rs +++ b/rustfs/src/storage/options.rs @@ -19,9 +19,10 @@ use http::{HeaderMap, HeaderValue}; use rustfs_utils::http::{ AMZ_BUCKET_REPLICATION_STATUS, SUFFIX_FORCE_DELETE, SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, SUFFIX_REPLICATION_ACTUAL_OBJECT_SIZE, SUFFIX_REPLICATION_SSEC_CRC, - SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_ETAG, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_REPLICATION_LEGALHOLD_TIMESTAMP, - SUFFIX_SOURCE_REPLICATION_REQUEST, SUFFIX_SOURCE_REPLICATION_RETENTION_TIMESTAMP, - SUFFIX_SOURCE_REPLICATION_TAGGING_TIMESTAMP, SUFFIX_SOURCE_VERSION_ID, SUFFIX_TAGGING_TIMESTAMP, get_header, + SUFFIX_SOURCE_DELETEMARKER, SUFFIX_SOURCE_ETAG, SUFFIX_SOURCE_MTIME, SUFFIX_SOURCE_PROXY_REQUEST, + SUFFIX_SOURCE_REPLICATION_LEGALHOLD_TIMESTAMP, SUFFIX_SOURCE_REPLICATION_REQUEST, + SUFFIX_SOURCE_REPLICATION_RETENTION_TIMESTAMP, SUFFIX_SOURCE_REPLICATION_TAGGING_TIMESTAMP, SUFFIX_SOURCE_VERSION_ID, + SUFFIX_TAGGING_TIMESTAMP, get_header, header_compat::{MINIO_ENCRYPTION_PREFIX, RUSTFS_ENCRYPTION_PREFIX}, insert_header_map, insert_str, metadata_compat::{MINIO_INTERNAL_PREFIX, RUSTFS_INTERNAL_PREFIX}, @@ -276,6 +277,19 @@ pub async fn get_opts( // Background scanner still performs full integrity checks asynchronously. opts.skip_verify_bitrot = get_skip_verify_bitrot(); + // Anti-loop markers for the replication read proxy + // (`{x-rustfs-,x-minio-}source-proxy-request` header family). + // MinIO semantics: the header being PRESENT at all (`ProxyHeaderSet`) + // disables proxying, whatever its value — a peer's replication worker + // sends "false" on its convergence HEADs so the receiver answers locally + // instead of proxying the miss back (a proxied echo would fake + // convergence and the object would never replicate). Deliberately not + // gated on replication authorization: the header only disables proxying + // (it grants nothing). + let proxy_header = get_header(headers, SUFFIX_SOURCE_PROXY_REQUEST); + opts.proxy_header_set = proxy_header.is_some(); + opts.proxy_request = proxy_header.map(|v| v.as_ref() == "true").unwrap_or_default(); + fill_conditional_writes_opts_from_header(headers, &mut opts)?; Ok(opts) @@ -2544,4 +2558,80 @@ mod tests { } } } + + /// The replication read-proxy anti-loop markers must be honored under + /// both interop prefixes (a MinIO peer sends x-minio-, a RustFS peer + /// sends both). `proxy_request` is set only for the literal value + /// "true", while `proxy_header_set` (MinIO `ProxyHeaderSet`) is set by + /// the header's mere presence — "false" (the replication worker's + /// convergence-HEAD marker) and arbitrary values included — so the + /// selector refuses to proxy either way. + #[tokio::test] + async fn test_get_opts_parses_source_proxy_request_under_both_prefixes() { + for header_name in ["x-rustfs-source-proxy-request", "x-minio-source-proxy-request"] { + let mut headers = HeaderMap::new(); + headers.insert(header_name, HeaderValue::from_static("true")); + let opts = get_opts("test-bucket", "test-object", None, None, &headers) + .await + .expect("get_opts should succeed"); + assert!(opts.proxy_request, "{header_name} must set opts.proxy_request"); + assert!(opts.proxy_header_set, "{header_name} must set opts.proxy_header_set"); + } + + let opts = get_opts("test-bucket", "test-object", None, None, &HeaderMap::new()) + .await + .expect("get_opts should succeed"); + assert!(!opts.proxy_request, "absent header must leave proxy_request off"); + assert!(!opts.proxy_header_set, "absent header must leave proxy_header_set off"); + + for (header_name, value) in [ + ("x-minio-source-proxy-request", "false"), + ("x-rustfs-source-proxy-request", "false"), + ("x-minio-source-proxy-request", "anything-else"), + ] { + let mut headers = HeaderMap::new(); + headers.insert(header_name, HeaderValue::from_static(value)); + let opts = get_opts("test-bucket", "test-object", None, None, &headers) + .await + .expect("get_opts should succeed"); + assert!(!opts.proxy_request, "{header_name}: non-'true' value must leave proxy_request off"); + assert!( + opts.proxy_header_set, + "{header_name}: value {value:?} must still set proxy_header_set (presence disables proxying)" + ); + } + } + + /// Pin that the source-proxy-request transport family cannot be + /// materialized as bare stored metadata via an `x-*-meta-` disguise: the + /// reserved-key namespacing (`x-rustfs-source-` / `x-minio-source-` + /// prefixes in `is_reserved_user_metadata_key`) must keep covering it. + #[test] + fn test_source_proxy_request_family_is_reserved_user_metadata() { + let mut headers = HeaderMap::new(); + headers.insert("x-amz-meta-x-minio-source-proxy-request", HeaderValue::from_static("true")); + headers.insert("x-rustfs-meta-x-rustfs-source-proxy-request", HeaderValue::from_static("true")); + // The bare transport header itself is not a user-metadata prefix and + // must never land in stored metadata at all. + headers.insert("x-minio-source-proxy-request", HeaderValue::from_static("true")); + + let metadata = extract_metadata(&headers); + + assert!( + !metadata.contains_key("x-minio-source-proxy-request"), + "bare source-proxy-request key must not be storable: {metadata:?}" + ); + assert!( + !metadata.contains_key("x-rustfs-source-proxy-request"), + "bare source-proxy-request key must not be storable: {metadata:?}" + ); + assert!( + metadata.contains_key("x-amz-meta-x-minio-source-proxy-request"), + "disguised key must be namespaced back under x-amz-meta-: {metadata:?}" + ); + assert!( + metadata.contains_key("x-amz-meta-x-rustfs-source-proxy-request"), + "disguised key must be namespaced back under x-amz-meta-: {metadata:?}" + ); + } } diff --git a/rustfs/src/storage/storage_api.rs b/rustfs/src/storage/storage_api.rs index be9548619..534269546 100644 --- a/rustfs/src/storage/storage_api.rs +++ b/rustfs/src/storage/storage_api.rs @@ -809,6 +809,10 @@ impl StorageReplicationStatsHandle { proxy_head_failed: metrics.proxied.head_failed, proxy_put_tag_total: metrics.proxied.put_tag_total, proxy_put_tag_failed: metrics.proxied.put_tag_failed, + proxy_get_tag_total: metrics.proxied.get_tag_total, + proxy_get_tag_failed: metrics.proxied.get_tag_failed, + proxy_delete_tag_total: metrics.proxied.delete_tag_total, + proxy_delete_tag_failed: metrics.proxied.delete_tag_failed, replica_size: metrics.replica_size, replica_count: metrics.replica_count, } @@ -845,6 +849,10 @@ pub(crate) struct ReplicationSiteMetricsSnapshot { pub(crate) proxy_head_failed: i64, pub(crate) proxy_put_tag_total: i64, pub(crate) proxy_put_tag_failed: i64, + pub(crate) proxy_get_tag_total: i64, + pub(crate) proxy_get_tag_failed: i64, + pub(crate) proxy_delete_tag_total: i64, + pub(crate) proxy_delete_tag_failed: i64, pub(crate) replica_size: i64, pub(crate) replica_count: i64, } @@ -1491,12 +1499,6 @@ pub(crate) async fn get_bucket_object_lock_config( ecstore_bucket::metadata_sys::get_object_lock_config(bucket).await } -pub(crate) async fn get_bucket_replication_config( - bucket: &str, -) -> Result<(s3s::dto::ReplicationConfiguration, time::OffsetDateTime)> { - ecstore_bucket::metadata_sys::get_replication_config(bucket).await -} - pub(crate) async fn persist_force_delete_intent( api: Arc, entry: ecstore_bucket::replication::MrfReplicateEntry, @@ -1842,18 +1844,6 @@ pub(crate) async fn find_local_disk_by_ref(disk_ref: &str) -> Option ecstore_storage::find_local_disk_by_ref(disk_ref).await } -pub(crate) trait StorageReplicationConfigExt { - fn has_active_rules(&self, prefix: &str, recursive: bool) -> bool; -} - -impl StorageReplicationConfigExt for s3s::dto::ReplicationConfiguration { - fn has_active_rules(&self, prefix: &str, recursive: bool) -> bool { - ::has_active_rules( - self, prefix, recursive, - ) - } -} - pub(crate) trait StorageVersioningConfigExt { fn enabled(&self) -> bool; } diff --git a/scripts/check_architecture_migration_rules.sh b/scripts/check_architecture_migration_rules.sh index 08c79d4ef..d510d49a6 100755 --- a/scripts/check_architecture_migration_rules.sh +++ b/scripts/check_architecture_migration_rules.sh @@ -242,6 +242,7 @@ REPLICATION_RESYNC_CONTRACT_BACKSLIDE_HITS_FILE="${TMP_DIR}/replication_resync_c REPLICATION_RESYNC_BOUNDARY_BYPASS_HITS_FILE="${TMP_DIR}/replication_resync_boundary_bypass_hits.txt" REPLICATION_OBJECT_DECISION_BOUNDARY_BYPASS_HITS_FILE="${TMP_DIR}/replication_object_decision_boundary_bypass_hits.txt" REPLICATION_OBJECT_COMPARE_CONTRACT_BACKSLIDE_HITS_FILE="${TMP_DIR}/replication_object_compare_contract_backslide_hits.txt" +REPLICATION_M2_DECISION_CONTRACT_BACKSLIDE_HITS_FILE="${TMP_DIR}/replication_m2_decision_contract_backslide_hits.txt" REPLICATION_MRF_WIRE_FORMAT_BACKSLIDE_HITS_FILE="${TMP_DIR}/replication_mrf_wire_format_backslide_hits.txt" STORAGE_REPLICATION_HANDLE_BOUNDARY_BYPASS_HITS_FILE="${TMP_DIR}/storage_replication_handle_boundary_bypass_hits.txt" STORAGE_REPLICATION_CRATE_BYPASS_HITS_FILE="${TMP_DIR}/storage_replication_crate_bypass_hits.txt" @@ -2981,7 +2982,7 @@ fi ( cd "$ROOT_DIR" replication_resync_status=0 - rg -n --with-filename '^\s*(?:pub(?:\([^)]*\))?\s+)?(?:(?:struct|enum)\s+(?:ResyncOpts|TargetReplicationResyncStatus|BucketReplicationResyncStatus|ResyncStatusType)|fn\s+(?:resync_state_accepts_update|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch))\b' \ + rg -n --with-filename '^\s*(?:pub(?:\([^)]*\))?\s+)?(?:(?:struct|enum)\s+(?:ResyncOpts|TargetReplicationResyncStatus|BucketReplicationResyncStatus|ResyncStatusType)|fn\s+(?:resync_state_accepts_update|resync_status_duration|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch))\b' \ crates/ecstore/src/bucket/replication \ --glob '*.rs' >"$REPLICATION_RESYNC_CONTRACT_BACKSLIDE_HITS_FILE" || replication_resync_status=$? if [[ "$replication_resync_status" -ne 0 && "$replication_resync_status" -ne 1 ]]; then @@ -2996,7 +2997,7 @@ fi ( cd "$ROOT_DIR" replication_resync_boundary_status=0 - rg -n --with-filename 'rustfs_replication::(resync::(RESYNC_META_FORMAT|RESYNC_META_VERSION|WIRE_ZERO_TIME_UNIX)|mrf::(MRF_META_FORMAT|MRF_META_VERSION)|(encode_resync_file|decode_resync_file|encode_mrf_file|decode_mrf_file|BucketReplicationResyncStatus|ResyncOpts|TargetReplicationResyncStatus|resync_state_accepts_update|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch)\b)' \ + rg -n --with-filename 'rustfs_replication::(resync::(RESYNC_META_FORMAT|RESYNC_META_VERSION|WIRE_ZERO_TIME_UNIX)|mrf::(MRF_META_FORMAT|MRF_META_VERSION)|(encode_resync_file|decode_resync_file|encode_mrf_file|decode_mrf_file|BucketReplicationResyncStatus|ResyncOpts|TargetReplicationResyncStatus|resync_state_accepts_update|resync_status_duration|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch)\b)' \ crates/ecstore/src/bucket/replication \ --glob '*.rs' \ --glob '!replication_resync_boundary.rs' >"$REPLICATION_RESYNC_BOUNDARY_BYPASS_HITS_FILE" || replication_resync_boundary_status=$? @@ -3005,7 +3006,7 @@ fi fi replication_resync_boundary_grouped_status=0 - rg -n -U --with-filename 'use\s+rustfs_replication::\{[^}]*\b(encode_resync_file|decode_resync_file|encode_mrf_file|decode_mrf_file|BucketReplicationResyncStatus|ResyncOpts|TargetReplicationResyncStatus|resync_state_accepts_update|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch)\b' \ + rg -n -U --with-filename 'use\s+rustfs_replication::\{[^}]*\b(encode_resync_file|decode_resync_file|encode_mrf_file|decode_mrf_file|BucketReplicationResyncStatus|ResyncOpts|TargetReplicationResyncStatus|resync_state_accepts_update|resync_status_duration|should_count_head_proxy_failure|should_auto_resume_resync|is_version_id_mismatch)\b' \ crates/ecstore/src/bucket/replication \ --glob '*.rs' \ --glob '!replication_resync_boundary.rs' >>"$REPLICATION_RESYNC_BOUNDARY_BYPASS_HITS_FILE" || replication_resync_boundary_grouped_status=$? @@ -3021,7 +3022,7 @@ fi ( cd "$ROOT_DIR" replication_object_decision_boundary_status=0 - rg -n --with-filename 'rustfs_replication::(MustReplicateOptions|ReplicationDeleteSource|ReplicationMultipartPartInput|ReplicationResyncTargetObject|delete_replication_missing_source_decision|delete_replication_object_opts|heal_uses_delete_replication_path|is_retryable_delete_replication_head_error|is_version_delete_replication|replication_etags_match|replication_multipart_complete_actual_size|replication_multipart_part_plan|resync_target_for_object|should_retry_delete_marker_purge)\b' \ + rg -n --with-filename 'rustfs_replication::(MustReplicateOptions|ReplicationDeleteSource|ReplicationMultipartPartInput|ReplicationResyncTargetObject|delete_marker_purge_mrf_entry|delete_marker_purge_version_id|delete_replication_missing_source_decision|delete_replication_object_opts|heal_uses_delete_replication_path|is_retryable_delete_replication_head_error|is_version_delete_replication|replicate_delete_outcome|replication_etags_match|replication_multipart_complete_actual_size|replication_multipart_part_plan|resync_existing_delete_replication_info|resync_target_for_object|should_retry_delete_marker_purge|target_delete_version_id)\b' \ crates/ecstore/src/bucket/replication \ --glob '*.rs' \ --glob '!replication_object_decision_boundary.rs' >"$REPLICATION_OBJECT_DECISION_BOUNDARY_BYPASS_HITS_FILE" || replication_object_decision_boundary_status=$? @@ -3030,7 +3031,7 @@ fi fi replication_object_decision_boundary_grouped_status=0 - rg -n -U --with-filename 'use\s+rustfs_replication::\{[^}]*\b(MustReplicateOptions|ReplicationDeleteSource|ReplicationMultipartPartInput|ReplicationResyncTargetObject|delete_replication_missing_source_decision|delete_replication_object_opts|heal_uses_delete_replication_path|is_retryable_delete_replication_head_error|is_version_delete_replication|replication_etags_match|replication_multipart_complete_actual_size|replication_multipart_part_plan|resync_target_for_object|should_retry_delete_marker_purge)\b' \ + rg -n -U --with-filename 'use\s+rustfs_replication::\{[^}]*\b(MustReplicateOptions|ReplicationDeleteSource|ReplicationMultipartPartInput|ReplicationResyncTargetObject|delete_marker_purge_mrf_entry|delete_marker_purge_version_id|delete_replication_missing_source_decision|delete_replication_object_opts|heal_uses_delete_replication_path|is_retryable_delete_replication_head_error|is_version_delete_replication|replicate_delete_outcome|replication_etags_match|replication_multipart_complete_actual_size|replication_multipart_part_plan|resync_existing_delete_replication_info|resync_target_for_object|should_retry_delete_marker_purge|target_delete_version_id)\b' \ crates/ecstore/src/bucket/replication \ --glob '*.rs' \ --glob '!replication_object_decision_boundary.rs' >>"$REPLICATION_OBJECT_DECISION_BOUNDARY_BYPASS_HITS_FILE" || replication_object_decision_boundary_grouped_status=$? @@ -3058,6 +3059,25 @@ if [[ -s "$REPLICATION_OBJECT_COMPARE_CONTRACT_BACKSLIDE_HITS_FILE" ]]; then report_failure "replication object comparison contracts must stay in crates/replication: $(paste -sd '; ' "$REPLICATION_OBJECT_COMPARE_CONTRACT_BACKSLIDE_HITS_FILE")" fi +( + cd "$ROOT_DIR" + replication_m2_decision_status=0 + # M2-moved pure decision contracts must not be redefined in ECStore. The + # ssec_passthrough_evidence_present name is deliberately absent: ECStore + # keeps a thin HeadObjectOutput adapter under that name in + # replication_target_boundary.rs which delegates to the crate-owned pure fn. + rg -n --with-filename '^\s*(?:pub(?:\([^)]*\))?\s+)?(?:enum\s+(?:SsecPassthroughCapability|SsecPassthroughGate)|fn\s+(?:replicate_delete_outcome|target_delete_version_id|delete_marker_purge_version_id|delete_marker_purge_mrf_entry|resync_existing_delete_replication_info|version_identity_drifted|is_replication_target_offline_error|ssec_passthrough_gate))\b' \ + crates/ecstore/src/bucket/replication \ + --glob '*.rs' >"$REPLICATION_M2_DECISION_CONTRACT_BACKSLIDE_HITS_FILE" || replication_m2_decision_status=$? + if [[ "$replication_m2_decision_status" -ne 0 && "$replication_m2_decision_status" -ne 1 ]]; then + exit "$replication_m2_decision_status" + fi +) + +if [[ -s "$REPLICATION_M2_DECISION_CONTRACT_BACKSLIDE_HITS_FILE" ]]; then + report_failure "M2-moved replication decision contracts must stay in crates/replication: $(paste -sd '; ' "$REPLICATION_M2_DECISION_CONTRACT_BACKSLIDE_HITS_FILE")" +fi + ( cd "$ROOT_DIR" rg -n --with-filename 'rmp_serde::(to_vec_named|from_slice)|LittleEndian::(write_u16|read_u16)|const\s+MRF_META_(FORMAT|VERSION):\s+u16\s*=\s*1\b' \ @@ -3728,6 +3748,12 @@ fi crates/ecstore/src/bucket/replication \ --glob '*.rs' | rg -v '^crates/ecstore/src/bucket/replication/replication_target_boundary\.rs:' || true + # M2-moved target decision contracts (pure gate family, offline classifier, + # version identity drift) route through the target boundary only. + rg -n -U --with-filename 'rustfs_replication::\{[^;]*\b(?:SsecPassthroughCapability|SsecPassthroughGate|is_replication_target_offline_error|ssec_passthrough_gate|ssec_passthrough_evidence_present|version_identity_drifted)\b|rustfs_replication::(?:SsecPassthroughCapability|SsecPassthroughGate|is_replication_target_offline_error|ssec_passthrough_gate|ssec_passthrough_evidence_present|version_identity_drifted)\b' \ + crates/ecstore/src/bucket/replication \ + --glob '*.rs' | + rg -v '^crates/ecstore/src/bucket/replication/replication_target_boundary\.rs:' || true ) >"$REPLICATION_TARGET_BOUNDARY_BYPASS_HITS_FILE" if [[ -s "$REPLICATION_TARGET_BOUNDARY_BYPASS_HITS_FILE" ]]; then From 8315c23d49b4b0517d9db4643793d501dd6bc619 Mon Sep 17 00:00:00 2001 From: Zhengchao An Date: Tue, 18 Aug 2026 21:46:00 +0800 Subject: [PATCH 20/24] test(kms): move the Vault KV2 doc guard into check_fips_wording.sh (#6215) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit test(kms): move the Vault KV2 Transit-wrapping doc guard into check_fips_wording.sh `test_vault_kv2_sources_do_not_claim_transit_wrapping` asserted that four `include_str!`-pinned files never describe the Vault KV2 backend as wrapping key material through Vault's Transit engine. The invariant is a documentation-claim invariant with no behavioral twin by construction, and the test form was weak in both directions: it saw only four files (the same prose in a fifth file passed silently) and it stopped compiling — rather than reporting a violation — as soon as one of them was renamed. Move the four literals verbatim into `scripts/check_fips_wording.sh`, which already guards the adjacent cryptographic over-claim class (unsupported FIPS validation wording) and is anchored to the same policy document. The guard now greps every file under `crates/kms` for the same four case-sensitive literals and separately reports a moved pinned source instead of failing to build. `check_fips_wording.sh` previously ran only in `make pre-commit` / `pre-pr`, so wire it into the Quick Checks job of both CI workflows to keep the invariant's failure visibility at least as strong as the deleted test's. --- .config/make/lint-fmt.mak | 4 +- .github/workflows/ci-docs-only.yml | 3 + .github/workflows/ci.yml | 3 + crates/kms/src/config.rs | 28 +----- .../kms-cryptographic-compliance.md | 2 + scripts/check_fips_wording.sh | 87 +++++++++++++++++-- 6 files changed, 92 insertions(+), 35 deletions(-) diff --git a/.config/make/lint-fmt.mak b/.config/make/lint-fmt.mak index f473845e2..46f659440 100644 --- a/.config/make/lint-fmt.mak +++ b/.config/make/lint-fmt.mak @@ -66,8 +66,8 @@ s3s-footprint-check: ## Check the s3s dependency footprint ratchet stays frozen ./scripts/check_s3s_footprint.sh .PHONY: fips-wording-check -fips-wording-check: ## Check outward docs do not make unsupported FIPS claims - @echo "📣 Checking FIPS wording guard..." +fips-wording-check: ## Check docs and crates/kms do not over-claim crypto capabilities + @echo "📣 Checking cryptographic capability wording guard..." ./scripts/check_fips_wording.sh .PHONY: log-analyzer-rules-check diff --git a/.github/workflows/ci-docs-only.yml b/.github/workflows/ci-docs-only.yml index 17eca6cba..a156a61b9 100644 --- a/.github/workflows/ci-docs-only.yml +++ b/.github/workflows/ci-docs-only.yml @@ -117,6 +117,9 @@ jobs: - name: Check s3s footprint ratchet run: ./scripts/check_s3s_footprint.sh + - name: Check cryptographic capability wording + run: ./scripts/check_fips_wording.sh + - name: Check no planning docs committed run: ./scripts/check_no_planning_docs.sh diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index 1988cb54d..412145661 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -152,6 +152,9 @@ jobs: - name: Check s3s footprint ratchet run: ./scripts/check_s3s_footprint.sh + - name: Check cryptographic capability wording + run: ./scripts/check_fips_wording.sh + - name: Check no planning docs committed run: ./scripts/check_no_planning_docs.sh diff --git a/crates/kms/src/config.rs b/crates/kms/src/config.rs index bd6c4a0e1..6e8cc1d06 100644 --- a/crates/kms/src/config.rs +++ b/crates/kms/src/config.rs @@ -1729,30 +1729,10 @@ mod tests { assert!(config.validate().is_ok(), "deprecated mount_path must not be required"); } - #[test] - fn test_vault_kv2_sources_do_not_claim_transit_wrapping() { - let sources = [ - ("config.rs", include_str!("config.rs")), - ("api_types.rs", include_str!("api_types.rs")), - ("backends/vault.rs", include_str!("backends/vault.rs")), - ("lib.rs", include_str!("lib.rs")), - ]; - // Assemble the needles at runtime so this guard does not match its own source. - let needles = [ - format!("wrapping via {}", "Transit"), - format!("KV v2 + {}", "Transit"), - format!("KV2+{}", "Transit"), - format!("you would use Vault's {} engine", "transit"), - ]; - for (name, source) in sources { - for needle in &needles { - assert!( - !source.contains(needle.as_str()), - "{name} still describes the Vault KV2 backend with `{needle}`" - ); - } - } - } + // The "VaultKv2 must not claim Transit wrapping" documentation-claim + // invariant is enforced by scripts/check_fips_wording.sh, which scans every + // file in crates/kms rather than a fixed include_str! list + // (rustfs/backlog#1884). #[test] fn test_legacy_persisted_vault_transit_config_uses_metadata_defaults() { diff --git a/docs/operations/kms-cryptographic-compliance.md b/docs/operations/kms-cryptographic-compliance.md index c9ce1a4d4..08dd593e3 100644 --- a/docs/operations/kms-cryptographic-compliance.md +++ b/docs/operations/kms-cryptographic-compliance.md @@ -53,6 +53,8 @@ Suggested boilerplate when the topic cannot be avoided: `README.md` and `CHANGELOG.md` currently contain no FIPS-related wording; `scripts/check_fips_wording.sh` is the grep guard for that public baseline. Any future occurrence of the banned strings in either file should be treated as a defect and either removed or brought under the qualifier rule above. This document intentionally contains the terminology needed to define the policy and is not part of that narrow outward-material scan. +The same script carries a second block for the adjacent over-claim: no file under `crates/kms` may describe the Vault KV2 backend as wrapping key material through Vault's Transit engine. `KmsBackend::VaultKv2` stores RustFS-wrapped key material in Vault's KV v2 engine and never calls Transit, so that wording would tell an operator their key material is cryptographically isolated inside Vault when it is not. Use the `VaultTransit` backend when that isolation is the requirement. + ## The `rustfs-crypto` `fips` feature: what it actually does `crates/crypto/Cargo.toml` declares `default = ["crypto", "fips"]`, so the feature is on in every normal build. Its entire effect is **which algorithm the write path selects**; the implementation is RustCrypto either way. diff --git a/scripts/check_fips_wording.sh b/scripts/check_fips_wording.sh index 49027119c..6221f6d4f 100755 --- a/scripts/check_fips_wording.sh +++ b/scripts/check_fips_wording.sh @@ -1,10 +1,30 @@ #!/usr/bin/env bash set -euo pipefail -# Guard: outward README and CHANGELOG material must not make an unsupported -# FIPS validation or certification claim. The detailed policy and permitted -# qualifiers live in docs/operations/kms-cryptographic-compliance.md; this -# check intentionally scans only the two public project-facing documents. +# Guard: cryptographic capability wording must not over-claim what RustFS +# actually does. Two independent blocks, both anchored to the policy in +# docs/operations/kms-cryptographic-compliance.md: +# +# 1. Outward README and CHANGELOG material must not make an unsupported +# FIPS validation or certification claim. This block intentionally scans +# only the two public project-facing documents; the permitted qualifiers +# live in the policy document. +# +# 2. Nothing in crates/kms may describe the Vault KV2 backend as wrapping +# key material through Vault's Transit engine. `KmsBackend::VaultKv2` +# stores RustFS-wrapped key material in Vault's KV v2 engine and never +# calls Transit (see crates/kms/src/config.rs and +# docs/operations/kms-backend-security.md), so such prose tells operators +# their key material is cryptographically isolated inside Vault when it is +# not. +# +# Block 2 replaces the unit test `test_vault_kv2_sources_do_not_claim_transit_wrapping` +# that used to live in crates/kms/src/config.rs (rustfs/backlog#1884). The +# invariant is a documentation-claim invariant, so it has no behavioral twin by +# construction and belongs in a wording guard rather than in a test. The test +# could only see four `include_str!`-pinned files and stopped compiling — +# rather than reporting a violation — the moment one of them was renamed; this +# block scans every file in the crate and reports a rename explicitly. SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" ROOT_DIR="${CHECK_FIPS_WORDING_ROOT:-$(cd "${SCRIPT_DIR}/.." && pwd)}" @@ -21,12 +41,34 @@ FORBIDDEN_PATTERNS=( '(meets|satisfies)[[:space:]]+FIPS' ) -status=0 +KMS_CRATE_DIR="crates/kms" + +# The four files the retired unit test pinned with include_str!. They stay +# listed so that moving one out of crates/kms is reported here instead of +# silently shrinking the scan; the scan itself is not limited to them. +KMS_PINNED_SOURCES=( + "crates/kms/src/config.rs" + "crates/kms/src/api_types.rs" + "crates/kms/src/backends/vault.rs" + "crates/kms/src/lib.rs" +) + +# Literal, case-sensitive, and byte-for-byte the needles the retired test built +# at runtime via format!("wrapping via {}", "Transit") and friends. +KMS_VAULT_KV2_FORBIDDEN=( + 'wrapping via Transit' + 'KV v2 + Transit' + 'KV2+Transit' + "you would use Vault's transit engine" +) + +fips_status=0 +kms_status=0 for target in "${TARGETS[@]}"; do if [[ ! -f "$target" ]]; then printf 'FIPS wording guard failed: %s is missing\n' "$target" >&2 - status=1 + fips_status=1 continue fi @@ -35,14 +77,41 @@ for target in "${TARGETS[@]}"; do if [[ -n "$matches" ]]; then printf 'FIPS wording guard failed: forbidden pattern /%s/ in %s:\n%s\n' \ "$pattern" "$target" "$matches" >&2 - status=1 + fips_status=1 fi done done -if [[ "$status" -ne 0 ]]; then +for source in "${KMS_PINNED_SOURCES[@]}"; do + if [[ ! -f "$source" ]]; then + printf 'KMS wording guard failed: %s is missing; update KMS_PINNED_SOURCES in scripts/check_fips_wording.sh after moving it\n' \ + "$source" >&2 + kms_status=1 + fi +done + +if [[ -d "$KMS_CRATE_DIR" ]]; then + for pattern in "${KMS_VAULT_KV2_FORBIDDEN[@]}"; do + matches="$(grep -r -F -n -- "$pattern" "$KMS_CRATE_DIR" || true)" + if [[ -n "$matches" ]]; then + printf 'KMS wording guard failed: forbidden Vault KV2 claim "%s" in %s:\n%s\n' \ + "$pattern" "$KMS_CRATE_DIR" "$matches" >&2 + kms_status=1 + fi + done +fi + +if [[ "$fips_status" -ne 0 ]]; then printf 'Remove unsupported FIPS validation wording from README.md or CHANGELOG.md.\n' >&2 - exit "$status" +fi + +if [[ "$kms_status" -ne 0 ]]; then + printf 'The Vault KV2 backend does not wrap key material through Vault Transit; fix the wording in crates/kms.\n' >&2 +fi + +if [[ "$fips_status" -ne 0 || "$kms_status" -ne 0 ]]; then + exit 1 fi printf 'FIPS wording guard passed (README.md and CHANGELOG.md contain no forbidden claims).\n' +printf 'KMS wording guard passed (crates/kms claims no Vault KV2 Transit wrapping).\n' From 1d056d76055e0fb6fd72f9cb565f97b1acbe3dde Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=94=90=E5=B0=8F=E9=B8=AD?= Date: Tue, 18 Aug 2026 21:46:16 +0800 Subject: [PATCH 21/24] test(e2e): pin bounded physical reads for compressed multipart range GETs (#6167) Byte-exactness tests stay green if the compressed range seek regresses into decoding from byte zero: the returned bytes are still correct and only the read amplification explodes. Assert the cost side as well. The observation reuses rustfs_io_get_object_shard_read_observed_bytes_total, already emitted per shard read by the erasure layer, so no production code is instrumented. The OTLP collector learns to accumulate a second counter, keyed by its path/role/outcome labels rather than by data-point position, which is not stable across exports. Two failure modes the assertions guard against: - With RUSTFS_OBS_METER_INTERVAL=1, treating one unchanged sample as settled measures a delta of zero, because the range read's counter has not been exported yet. Settling now requires several consecutive equal samples. - An upper bound alone passes vacuously on a zero delta, so a lower bound turns "measured nothing" into a failure instead of a green run. Refs rustfs/rustfs#5957, backlog#1848. --- .../src/inline_fast_path_cluster_test.rs | 173 ++++++++++++++++++ 1 file changed, 173 insertions(+) diff --git a/crates/e2e_test/src/inline_fast_path_cluster_test.rs b/crates/e2e_test/src/inline_fast_path_cluster_test.rs index e2890c4a8..f1cdc1acf 100644 --- a/crates/e2e_test/src/inline_fast_path_cluster_test.rs +++ b/crates/e2e_test/src/inline_fast_path_cluster_test.rs @@ -67,6 +67,9 @@ type MetricValues = Arc>>; const KIB: usize = 1024; const READER_PATH_COUNTER: &str = "rustfs_io_get_object_reader_path_by_size_total"; +/// Physical bytes the erasure layer pulled from disk, emitted per shard read by +/// `crates/ecstore/src/erasure/coding/decode.rs`. +const SHARD_READ_BYTES_COUNTER: &str = "rustfs_io_get_object_shard_read_observed_bytes_total"; const MSGPACK_JSON_DECODE_COUNTER: &str = "rustfs_system_network_internode_msgpack_json_decode_total"; const MSGPACK_JSON_FALLBACK_COUNTER: &str = "rustfs_system_network_internode_msgpack_json_fallback_total"; const MSGPACK_JSON_DECODE_ERROR_COUNTER: &str = "rustfs_system_network_internode_msgpack_json_decode_error_total"; @@ -146,6 +149,7 @@ struct OtlpMetricCollector { decode_values: MetricValues, fallback_values: MetricValues, decode_error_values: MetricValues, + shard_read_values: MetricValues, task: JoinHandle<()>, } @@ -157,10 +161,12 @@ impl OtlpMetricCollector { let decode_values = Arc::new(Mutex::new(BTreeMap::new())); let fallback_values = Arc::new(Mutex::new(BTreeMap::new())); let decode_error_values = Arc::new(Mutex::new(BTreeMap::new())); + let shard_read_values = Arc::new(Mutex::new(BTreeMap::new())); let task_values = values.clone(); let task_decode_values = decode_values.clone(); let task_fallback_values = fallback_values.clone(); let task_decode_error_values = decode_error_values.clone(); + let task_shard_read_values = shard_read_values.clone(); let task = tokio::spawn(async move { loop { let Ok((stream, _)) = listener.accept().await else { @@ -170,6 +176,7 @@ impl OtlpMetricCollector { let decode_values = task_decode_values.clone(); let fallback_values = task_fallback_values.clone(); let decode_error_values = task_decode_error_values.clone(); + let shard_read_values = task_shard_read_values.clone(); tokio::spawn(async move { let _ = hyper::server::conn::http1::Builder::new() .serve_connection( @@ -181,6 +188,7 @@ impl OtlpMetricCollector { decode_values.clone(), fallback_values.clone(), decode_error_values.clone(), + shard_read_values.clone(), ) }), ) @@ -194,10 +202,48 @@ impl OtlpMetricCollector { decode_values, fallback_values, decode_error_values, + shard_read_values, task, }) } + /// Total physical bytes read from disk across every shard-read label set. + async fn shard_read_bytes_total(&self) -> u64 { + self.shard_read_values + .lock() + .await + .values() + .map(|versions| versions.values().map(|(_, value)| *value).sum::()) + .sum() + } + + /// Waits until the shard-read counter stops advancing so a measurement window + /// is not polluted by exports still in flight. + /// + /// Requires several consecutive equal samples spanning more than one export + /// interval (`RUSTFS_OBS_METER_INTERVAL=1`): a single unchanged sample only + /// proves the latest export has not landed yet, which silently reads as "no + /// disk reads happened" and makes any upper-bound assertion vacuous. + async fn wait_for_shard_read_bytes_to_settle(&self) -> TestResult { + const REQUIRED_STABLE_SAMPLES: usize = 5; + let mut last = self.shard_read_bytes_total().await; + let mut stable = 0; + for _ in 0..60 { + sleep(Duration::from_millis(500)).await; + let current = self.shard_read_bytes_total().await; + if current == last { + stable += 1; + if stable >= REQUIRED_STABLE_SAMPLES { + return Ok(current); + } + } else { + stable = 0; + last = current; + } + } + Err("timed out waiting for shard-read byte counter to settle".into()) + } + async fn reader_path_total(&self, path: &str, object_class: &str, size_bucket: &str) -> u64 { self.reader_path_values(path, object_class, size_bucket).await.values().sum() } @@ -321,6 +367,7 @@ async fn handle_metric_export( decode_values: MetricValues, fallback_values: MetricValues, decode_error_values: MetricValues, + shard_read_values: MetricValues, ) -> Result>, Infallible> { if request.uri().path() != "/v1/metrics" { return Ok(response(StatusCode::NOT_FOUND)); @@ -354,7 +401,9 @@ async fn handle_metric_export( let mut decode_values = decode_values.lock().await; let mut fallback_values = fallback_values.lock().await; let mut decode_error_values = decode_error_values.lock().await; + let mut shard_read_values = shard_read_values.lock().await; record_reader_path_metrics(&export, &mut values); + record_shard_read_bytes_metrics(&export, &mut shard_read_values); record_msgpack_decode_metrics(&export, &mut decode_values); record_msgpack_fallback_metrics(&export, &mut fallback_values); record_msgpack_decode_error_metrics(&export, &mut decode_error_values); @@ -375,6 +424,50 @@ fn reader_path_metric_key(path: &str, object_class: &str, size_bucket: &str) -> format!("{path}\u{1f}{object_class}\u{1f}{size_bucket}") } +/// Accumulates `SHARD_READ_BYTES_COUNTER` across all label sets. Only the total +/// matters: it is the number of physical bytes the erasure layer actually pulled +/// from disk, which is what separates a bounded per-part read from a decode of +/// the whole object. +fn record_shard_read_bytes_metrics(export: &ExportMetricsServiceRequest, values: &mut BTreeMap) { + for resource_metrics in &export.resource_metrics { + for scope_metrics in &resource_metrics.scope_metrics { + for metric in &scope_metrics.metrics { + if metric.name != SHARD_READ_BYTES_COUNTER { + continue; + } + let Some(metric::Data::Sum(sum)) = &metric.data else { + continue; + }; + for point in &sum.data_points { + let Some(number_data_point::Value::AsInt(value)) = point.value.as_ref() else { + continue; + }; + let value = u64::try_from(*value).unwrap_or_default(); + // Keyed by labels, not by position: point order within an export + // is not guaranteed stable, so an index key would alias distinct + // series across batches. + let key = format!( + "{}\u{1f}{}\u{1f}{}", + attribute_string(&point.attributes, "path").unwrap_or_default(), + attribute_string(&point.attributes, "role").unwrap_or_default(), + attribute_string(&point.attributes, "outcome").unwrap_or_default(), + ); + values + .entry(key) + .or_default() + .entry(point.start_time_unix_nano) + .and_modify(|current| { + if point.time_unix_nano >= current.0 { + *current = (point.time_unix_nano, value); + } + }) + .or_insert((point.time_unix_nano, value)); + } + } + } + } +} + fn record_reader_path_metrics(export: &ExportMetricsServiceRequest, values: &mut BTreeMap) { for resource_metrics in &export.resource_metrics { for scope_metrics in &resource_metrics.scope_metrics { @@ -1864,6 +1957,86 @@ async fn four_node_multipart_disk_compression_roundtrip() -> TestResult { Ok(()) } +/// A tail range over a compressed multipart object must read only the physical +/// data it needs, not decode the object from byte zero. +/// +/// The byte-exactness tests around this one stay green even if the seek path +/// regresses into decoding from the start of the object: the bytes returned are +/// still correct, only the read amplification explodes. This asserts the cost +/// side, using `SHARD_READ_BYTES_COUNTER` — already emitted per shard read by the +/// erasure layer, so no production code is instrumented for the test. +/// +/// `get_compressed_offsets` skips whole preceding parts by their stored size and +/// then seeks inside the covering part via its compression index, so a bounded +/// read costs on the order of the covering part's block size against a ~5 MiB +/// object. +#[tokio::test] +#[serial] +async fn four_node_compressed_multipart_tail_range_reads_are_bounded() -> TestResult { + init_logging(); + + let collector = OtlpMetricCollector::start().await?; + let mut cluster = RustFSTestClusterEnvironment::new(4).await?; + configure_reader_metric_cluster(&mut cluster, &collector); + cluster.set_env("RUSTFS_COMPRESSION_ENABLED", "true"); + cluster.set_env("RUSTFS_COMPRESSION_MULTIPART_ENABLED", "true"); + cluster.start().await?; + + let bucket = "inline-multipart-compression-tail-range"; + cluster.create_test_bucket(bucket).await?; + let client = cluster.create_s3_client(0)?; + let key = "multipart/tail-range.txt"; + let (body, _second_part, etag) = put_two_part_multipart(&client, bucket, key).await?; + + // Establish that the object really took the compressed read path; otherwise a + // small delta below would only prove compression never happened. + assert_reader_path( + &collector, + &client, + ReaderPathExpectation::for_class(ReaderObject::new(bucket, key, &body, etag.as_deref(), None), LEGACY_DUPLEX, COMPRESSED), + ) + .await?; + + let baseline = collector.wait_for_shard_read_bytes_to_settle().await?; + + let tail_len = 4 * KIB; + let start = body.len() - tail_len; + let end = body.len() - 1; + let range = client + .get_object() + .bucket(bucket) + .key(key) + .range(format!("bytes={start}-{end}")) + .send() + .await?; + let tail = range.body.collect().await?.into_bytes(); + assert_eq!(tail.as_ref(), &body[start..], "tail range returned wrong bytes"); + + let after = collector.wait_for_shard_read_bytes_to_settle().await?; + let read_bytes = after.saturating_sub(baseline); + + // A zero delta means the window caught nothing — an unexported counter, or a + // read served without touching the erasure layer — which would make the upper + // bound vacuously true. Fail instead of passing blind. + assert!( + read_bytes > 0, + "no shard reads observed for the tail range; the budget assertion below would be vacuous" + ); + + // Part 1 alone is MPU_PART_1_SIZE, so a whole-object decode cannot come in + // under it. Half the logical size leaves generous headroom for erasure padding + // and unrelated background reads while still failing loudly on a full decode. + let budget = (body.len() / 2) as u64; + assert!( + read_bytes < budget, + "tail range read {read_bytes} physical bytes for a {tail_len}-byte range (budget {budget}, object {} bytes): \ + the read is not bounded to the covering part", + body.len() + ); + + Ok(()) +} + #[tokio::test] #[serial] async fn four_node_mixed_msgpack_compat_mode_preserves_fallback_controls() -> TestResult { From 91c97f3416b143d1e993f58c3ec7148abd92a00a Mon Sep 17 00:00:00 2001 From: GatewayJ <835269233@qq.com> Date: Tue, 18 Aug 2026 21:46:56 +0800 Subject: [PATCH 22/24] chore(deps): update s3s to upstream main (#6203) * deps: update s3s to upstream main * deps: refresh s3s upstream revision * deps: pin s3s to latest upstream main --------- Co-authored-by: houseme Co-authored-by: heihutu --- Cargo.lock | 2 +- Cargo.toml | 2 +- rustfs/src/app/metadata_route.rs | 24 ++++++++++++++++++++++++ 3 files changed, 26 insertions(+), 2 deletions(-) diff --git a/Cargo.lock b/Cargo.lock index db75a7992..c73ac8d64 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -10678,7 +10678,7 @@ checksum = "9774ba4a74de5f7b1c1451ed6cd5285a32eddb5cccb8cc655a4e50009e06477f" [[package]] name = "s3s" version = "0.14.1" -source = "git+https://github.com/rustfs/s3s.git?rev=ff8106f46098b60fce8ee88518f34974fbfbec4b#ff8106f46098b60fce8ee88518f34974fbfbec4b" +source = "git+https://github.com/rustfs/s3s.git?rev=d358a68783096df1db0c3e314127f2704603b29e#d358a68783096df1db0c3e314127f2704603b29e" dependencies = [ "arc-swap", "arrayvec", diff --git a/Cargo.toml b/Cargo.toml index f2080a15f..b6a37b028 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -290,7 +290,7 @@ rustify = { version = "0.7", default-features = false } rustix = { version = "1.1.4" } rust-embed = { version = "8.12.0" } rustc-hash = { version = "2.1.3" } -s3s = { git = "https://github.com/rustfs/s3s.git", rev = "ff8106f46098b60fce8ee88518f34974fbfbec4b" } +s3s = { git = "https://github.com/rustfs/s3s.git", rev = "d358a68783096df1db0c3e314127f2704603b29e" } serial_test = "4.0.1" shadow-rs = { default-features = false, version = "2.0.0" } siphasher = "1.0.3" diff --git a/rustfs/src/app/metadata_route.rs b/rustfs/src/app/metadata_route.rs index d6509f60e..3a49a41e7 100644 --- a/rustfs/src/app/metadata_route.rs +++ b/rustfs/src/app/metadata_route.rs @@ -349,6 +349,30 @@ mod tests { ); } + #[test] + fn metadata_operation_matches_unconfigured_host_fallbacks() { + let host = MultiDomain::new(["s3.example.com", "s3.example.com:9000"]).expect("valid test host domain"); + + let mut path_style_headers = HeaderMap::new(); + path_style_headers.insert(HOST, "localhost:9000".parse().expect("valid host header")); + let path_style = metadata_operation( + &Method::GET, + &uri("/path-bucket?list-type=2&metadata=true"), + &path_style_headers, + Some(&host), + ) + .expect("unmatched host with a port should use path-style routing"); + assert_eq!(path_style.operation, MetadataOperation::ListObjectsV2); + assert_eq!(path_style.bucket, "path-bucket"); + + let mut cname_headers = HeaderMap::new(); + cname_headers.insert(HOST, "cdn.example.org".parse().expect("valid host header")); + let cname = metadata_operation(&Method::GET, &uri("/?list-type=2&metadata=true"), &cname_headers, Some(&host)) + .expect("unmatched valid bucket host should use CNAME routing"); + assert_eq!(cname.operation, MetadataOperation::ListObjectsV2); + assert_eq!(cname.bucket, "cdn.example.org"); + } + #[test] fn list_objects_v2_input_parses_query_headers_and_decodes_bucket() { let mut headers = HeaderMap::new(); From eed0ca3612a530d750b7bd746ae71fef05b8fe4d Mon Sep 17 00:00:00 2001 From: houseme Date: Tue, 18 Aug 2026 22:40:43 +0800 Subject: [PATCH 23/24] perf(ecstore): validate local IO paths with openat2 (#6221) Use Linux openat2 with RESOLVE_BENEATH and RESOLVE_NO_SYMLINKS for LocalDisk I/O path validation while keeping the existing lstat walk as the public-path and unsupported-kernel fallback. Add focused regression coverage for traversal, symlink swaps, missing leaves, recreated parents, high-cardinality prefixes, final symlink leaves, and concurrent validation. Co-authored-by: heihutu --- crates/ecstore/src/disk/local.rs | 364 ++++++++++++++++++++++++++++++- 1 file changed, 356 insertions(+), 8 deletions(-) diff --git a/crates/ecstore/src/disk/local.rs b/crates/ecstore/src/disk/local.rs index a71f2c55b..273ada7d7 100644 --- a/crates/ecstore/src/disk/local.rs +++ b/crates/ecstore/src/disk/local.rs @@ -5618,11 +5618,37 @@ impl LocalDisk { } fn io_get_object_path(&self, bucket: &str, key: &str) -> Result { - local_disk_object_path(self.io_root(), bucket, key) + self.local_disk_object_path(self.io_root(), bucket, key) } fn io_get_bucket_path(&self, bucket: &str) -> Result { - local_disk_bucket_path(self.io_root(), bucket) + self.local_disk_bucket_path(self.io_root(), bucket) + } + + fn local_disk_object_path(&self, root: &Path, bucket: &str, key: &str) -> Result { + let (bucket_path, path) = build_local_disk_object_path(root, bucket, key); + #[cfg(target_os = "linux")] + { + check_local_disk_valid_object_path_at(root, &self.mount_lease, &bucket_path, &path)?; + } + #[cfg(not(target_os = "linux"))] + { + check_local_disk_valid_object_path(root, &bucket_path, &path)?; + } + Ok(path) + } + + fn local_disk_bucket_path(&self, root: &Path, bucket: &str) -> Result { + let bucket_path = build_local_disk_bucket_path(root, bucket); + #[cfg(target_os = "linux")] + { + check_local_disk_valid_path_at(root, &self.mount_lease, &bucket_path)?; + } + #[cfg(not(target_os = "linux"))] + { + check_local_disk_valid_path(root, &bucket_path)?; + } + Ok(bucket_path) } // Check if a path is valid @@ -5631,7 +5657,14 @@ impl LocalDisk { reason = "method wrapper over the live free function check_local_disk_valid_path; no caller in this port (backlog#1823)" )] fn check_valid_path>(&self, path: P) -> Result<()> { - check_local_disk_valid_path(self.io_root(), path) + #[cfg(target_os = "linux")] + { + check_local_disk_valid_path_at(self.io_root(), &self.mount_lease, path) + } + #[cfg(not(target_os = "linux"))] + { + check_local_disk_valid_path(self.io_root(), path) + } } #[allow( @@ -5639,7 +5672,14 @@ impl LocalDisk { reason = "method wrapper over the live free function reject_local_disk_symlink_components; no caller in this port (backlog#1823)" )] fn reject_symlink_components(&self, path: &Path) -> Result<()> { - reject_local_disk_symlink_components(self.io_root(), path) + #[cfg(target_os = "linux")] + { + reject_local_disk_symlink_components_at(self.io_root(), &self.mount_lease, path) + } + #[cfg(not(target_os = "linux"))] + { + reject_local_disk_symlink_components(self.io_root(), path) + } } // Batch path generation with single lock acquisition @@ -7323,29 +7363,54 @@ fn skip_access_checks(p: impl AsRef) -> bool { } fn local_disk_object_path(root: &Path, bucket: &str, key: &str) -> Result { + let (bucket_path, path) = build_local_disk_object_path(root, bucket, key); + check_local_disk_valid_object_path(root, &bucket_path, &path)?; + Ok(path) +} + +fn build_local_disk_object_path(root: &Path, bucket: &str, key: &str) -> (PathBuf, PathBuf) { let cache_key = if key.is_empty() { bucket.to_string() } else { path_join_buf(&[bucket, key]) }; + #[cfg(windows)] + let bucket_path = root.join(bucket.replace('/', "\\")); + #[cfg(not(windows))] + let bucket_path = root.join(bucket); + #[cfg(windows)] let path = root.join(cache_key.replace('/', "\\")); #[cfg(not(windows))] let path = root.join(cache_key); - check_local_disk_valid_path(root, &path)?; - Ok(path) + (bucket_path, path) } fn local_disk_bucket_path(root: &Path, bucket: &str) -> Result { + let bucket_path = build_local_disk_bucket_path(root, bucket); + check_local_disk_valid_path(root, &bucket_path)?; + Ok(bucket_path) +} + +fn build_local_disk_bucket_path(root: &Path, bucket: &str) -> PathBuf { #[cfg(windows)] let bucket_path = root.join(bucket.replace('/', "\\")); #[cfg(not(windows))] let bucket_path = root.join(bucket); - check_local_disk_valid_path(root, &bucket_path)?; - Ok(bucket_path) + bucket_path +} + +fn check_local_disk_valid_object_path(root: &Path, bucket_path: &Path, path: &Path) -> Result<()> { + let bucket_path = normalize_path_components(bucket_path); + let path = normalize_path_components(path); + if !bucket_path.starts_with(root) || !path.starts_with(&bucket_path) { + return Err(DiskError::InvalidPath); + } + + reject_local_disk_symlink_components(root, &path) } fn check_local_disk_valid_path(root: &Path, path: impl AsRef) -> Result<()> { @@ -7357,6 +7422,80 @@ fn check_local_disk_valid_path(root: &Path, path: impl AsRef) -> Result<() reject_local_disk_symlink_components(root, &path) } +#[cfg(target_os = "linux")] +fn check_local_disk_valid_object_path_at(root: &Path, root_fd: &std::fs::File, bucket_path: &Path, path: &Path) -> Result<()> { + let bucket_path = normalize_path_components(bucket_path); + let path = normalize_path_components(path); + if !bucket_path.starts_with(root) || !path.starts_with(&bucket_path) { + return Err(DiskError::InvalidPath); + } + + reject_local_disk_symlink_components_at(root, root_fd, &path) +} + +#[cfg(target_os = "linux")] +fn check_local_disk_valid_path_at(root: &Path, root_fd: &std::fs::File, path: impl AsRef) -> Result<()> { + let path = normalize_path_components(path); + if !path.starts_with(root) { + return Err(DiskError::InvalidPath); + } + + reject_local_disk_symlink_components_at(root, root_fd, &path) +} + +#[cfg(target_os = "linux")] +fn reject_local_disk_symlink_components_at(root: &Path, root_fd: &std::fs::File, path: &Path) -> Result<()> { + let relative = path.strip_prefix(root).map_err(|_| DiskError::InvalidPath)?; + match validate_existing_local_disk_prefix_at(root_fd, relative) { + Ok(()) => Ok(()), + Err(LocalDiskPathValidationAtError::Unsupported) => reject_local_disk_symlink_components(root, path), + Err(LocalDiskPathValidationAtError::InvalidPath) => Err(DiskError::InvalidPath), + Err(LocalDiskPathValidationAtError::Io(err)) => Err(to_file_error(err).into()), + } +} + +#[cfg(target_os = "linux")] +enum LocalDiskPathValidationAtError { + Unsupported, + InvalidPath, + Io(std::io::Error), +} + +#[cfg(target_os = "linux")] +fn validate_existing_local_disk_prefix_at( + root_fd: &std::fs::File, + relative: &Path, +) -> core::result::Result<(), LocalDiskPathValidationAtError> { + use rustix::fs::{Mode, OFlags, ResolveFlags, openat2}; + use rustix::io::Errno; + + if relative.as_os_str().is_empty() { + return Ok(()); + } + + let mut candidate = relative.to_path_buf(); + loop { + match openat2( + root_fd, + &candidate, + OFlags::PATH | OFlags::CLOEXEC, + Mode::empty(), + ResolveFlags::BENEATH | ResolveFlags::NO_SYMLINKS, + ) { + Ok(_) => return Ok(()), + Err(Errno::NOSYS) => return Err(LocalDiskPathValidationAtError::Unsupported), + Err(Errno::LOOP | Errno::XDEV) => return Err(LocalDiskPathValidationAtError::InvalidPath), + Err(Errno::NOENT) => { + let Some(parent) = candidate.parent().filter(|parent| !parent.as_os_str().is_empty()) else { + return Ok(()); + }; + candidate = parent.to_path_buf(); + } + Err(err) => return Err(LocalDiskPathValidationAtError::Io(err.into())), + } + } +} + fn reject_local_disk_symlink_components(root: &Path, path: &Path) -> Result<()> { let relative = path.strip_prefix(root).map_err(|_| DiskError::InvalidPath)?; let mut current = root.to_path_buf(); @@ -18068,6 +18207,22 @@ mod test { assert!(matches!(disk.get_bucket_path("escape-bucket"), Err(DiskError::InvalidPath))); } + #[cfg(unix)] + #[tokio::test] + async fn get_bucket_path_for_io_rejects_symlink_escape() { + use std::os::unix::fs::symlink; + + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let outside_dir = tempfile::tempdir().expect("outside temp dir should be created"); + let link_path = root_dir.path().join("escape-bucket"); + symlink(outside_dir.path(), &link_path).expect("bucket symlink should be created"); + + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + + assert!(matches!(disk.get_bucket_path_for_io("escape-bucket"), Err(DiskError::InvalidPath))); + } + #[cfg(unix)] #[tokio::test] async fn test_get_object_path_rejects_symlink_component_escape() { @@ -18087,6 +18242,199 @@ mod test { assert!(matches!(disk.get_object_path("bucket", "escape/object.txt"), Err(DiskError::InvalidPath))); } + #[cfg(unix)] + #[tokio::test] + async fn get_object_path_for_io_rejects_symlink_leaf() { + use std::os::unix::fs::symlink; + + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let outside_file = root_dir.path().join("outside-file"); + fs::write(&outside_file, b"outside") + .await + .expect("outside file should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + disk.make_volume("bucket").await.expect("bucket should be created"); + symlink(&outside_file, root_dir.path().join("bucket/object")).expect("object symlink should be created"); + + assert!(matches!(disk.get_object_path_for_io("bucket", "object"), Err(DiskError::InvalidPath))); + } + + #[tokio::test] + async fn get_object_path_rejects_key_traversal_out_of_bucket() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + + assert!(matches!(disk.get_object_path("bucket", "../outside"), Err(DiskError::InvalidPath))); + assert!(matches!( + disk.get_object_path("bucket", "prefix/../../outside"), + Err(DiskError::InvalidPath) + )); + } + + #[tokio::test] + async fn get_object_path_accepts_missing_leaf_under_existing_bucket() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + disk.make_volume("bucket").await.expect("bucket should be created"); + + let object_path = disk + .get_object_path("bucket", "missing-object") + .expect("missing leaf under a valid bucket should resolve"); + + assert_eq!(object_path, disk.root.join("bucket/missing-object")); + } + + #[tokio::test] + async fn get_object_path_for_io_rejects_key_traversal_out_of_bucket() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + + assert!(matches!(disk.get_object_path_for_io("bucket", "../outside"), Err(DiskError::InvalidPath))); + assert!(matches!( + disk.get_object_path_for_io("bucket", "prefix/../../outside"), + Err(DiskError::InvalidPath) + )); + } + + #[tokio::test] + async fn get_object_path_for_io_accepts_missing_leaf_under_existing_bucket() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + disk.make_volume("bucket").await.expect("bucket should be created"); + + let object_path = disk + .get_object_path_for_io("bucket", "missing-object") + .expect("missing leaf under a valid I/O bucket should resolve"); + + assert!(object_path.ends_with("bucket/missing-object")); + } + + #[cfg(unix)] + #[tokio::test] + async fn get_object_path_rejects_symlink_component_after_prior_valid_lookup() { + use std::os::unix::fs::symlink; + + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let outside_dir = tempfile::tempdir().expect("outside temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + let prefix = root_dir.path().join("bucket/prefix"); + fs::create_dir_all(&prefix).await.expect("prefix should be created"); + + disk.get_object_path("bucket", "prefix/object") + .expect("initial lookup should validate the real prefix"); + fs::remove_dir(&prefix).await.expect("prefix should be removable"); + symlink(outside_dir.path(), &prefix).expect("prefix should be replaced by a symlink"); + + assert!(matches!(disk.get_object_path("bucket", "prefix/object"), Err(DiskError::InvalidPath))); + } + + #[cfg(unix)] + #[tokio::test] + async fn get_object_path_for_io_rejects_symlink_component_after_prior_valid_lookup() { + use std::os::unix::fs::symlink; + + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let outside_dir = tempfile::tempdir().expect("outside temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + let prefix = root_dir.path().join("bucket/prefix"); + fs::create_dir_all(&prefix).await.expect("prefix should be created"); + + disk.get_object_path_for_io("bucket", "prefix/object") + .expect("initial I/O lookup should validate the real prefix"); + fs::remove_dir(&prefix).await.expect("prefix should be removable"); + symlink(outside_dir.path(), &prefix).expect("prefix should be replaced by a symlink"); + + assert!(matches!( + disk.get_object_path_for_io("bucket", "prefix/object"), + Err(DiskError::InvalidPath) + )); + } + + #[tokio::test] + async fn get_object_path_accepts_parent_recreated_after_prior_valid_lookup() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + let prefix = root_dir.path().join("bucket/prefix"); + fs::create_dir_all(&prefix).await.expect("prefix should be created"); + + disk.get_object_path("bucket", "prefix/object") + .expect("initial lookup should validate the real prefix"); + fs::remove_dir(&prefix).await.expect("prefix should be removable"); + fs::create_dir(&prefix).await.expect("prefix should be recreated"); + + let object_path = disk + .get_object_path("bucket", "prefix/object") + .expect("recreated non-symlink parent should validate"); + assert_eq!(object_path, disk.root.join("bucket/prefix/object")); + } + + #[tokio::test] + async fn get_object_path_for_io_accepts_parent_recreated_after_prior_valid_lookup() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + let prefix = root_dir.path().join("bucket/prefix"); + fs::create_dir_all(&prefix).await.expect("prefix should be created"); + + disk.get_object_path_for_io("bucket", "prefix/object") + .expect("initial I/O lookup should validate the real prefix"); + fs::remove_dir(&prefix).await.expect("prefix should be removable"); + fs::create_dir(&prefix).await.expect("prefix should be recreated"); + + let object_path = disk + .get_object_path_for_io("bucket", "prefix/object") + .expect("recreated non-symlink parent should validate for I/O"); + assert!(object_path.ends_with("bucket/prefix/object")); + } + + #[tokio::test] + async fn get_object_path_handles_many_unique_missing_prefixes_without_state_growth() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = LocalDisk::new(&endpoint, false).await.expect("local disk should be created"); + disk.make_volume("bucket").await.expect("bucket should be created"); + + for index in 0..5000 { + let object_path = disk + .get_object_path("bucket", &format!("prefix-{index}/object")) + .expect("unique missing prefix should validate without shared state"); + assert!(object_path.ends_with(format!("bucket/prefix-{index}/object"))); + } + } + + #[tokio::test] + async fn get_object_path_concurrent_validation_keeps_paths_under_bucket() { + let root_dir = tempfile::tempdir().expect("temp dir should be created"); + let endpoint = Endpoint::try_from(root_dir.path().to_string_lossy().as_ref()).expect("endpoint should parse"); + let disk = Arc::new(LocalDisk::new(&endpoint, false).await.expect("local disk should be created")); + disk.make_volume("bucket").await.expect("bucket should be created"); + let barrier = Arc::new(tokio::sync::Barrier::new(32)); + let mut tasks = Vec::with_capacity(32); + + for index in 0..32 { + let disk = disk.clone(); + let barrier = barrier.clone(); + tasks.push(tokio::spawn(async move { + barrier.wait().await; + disk.get_object_path("bucket", &format!("object-{index}")) + .expect("concurrent validation should resolve object path") + })); + } + + for task in tasks { + let object_path = task.await.expect("validation task should complete"); + assert!(object_path.starts_with(disk.root.join("bucket"))); + } + } + #[tokio::test] async fn test_local_disk_file_operations() { let test_dir = "./test_local_disk_file_ops"; From 905082893fc0cc1a80148c5975a98d656f45398d Mon Sep 17 00:00:00 2001 From: houseme Date: Tue, 18 Aug 2026 23:12:39 +0800 Subject: [PATCH 24/24] fix(e2e): import serial test attribute (#6222) --- crates/e2e_test/src/replication_extension_test.rs | 1 + 1 file changed, 1 insertion(+) diff --git a/crates/e2e_test/src/replication_extension_test.rs b/crates/e2e_test/src/replication_extension_test.rs index 8279479f1..b9645deba 100644 --- a/crates/e2e_test/src/replication_extension_test.rs +++ b/crates/e2e_test/src/replication_extension_test.rs @@ -60,6 +60,7 @@ use rustfs_signer::constants::UNSIGNED_PAYLOAD; use rustfs_signer::sign_v4; use s3s::Body; use s3s::header::X_AMZ_REPLICATION_STATUS; +use serial_test::serial; use sha2::{Digest, Sha256}; use std::collections::BTreeMap; use std::convert::Infallible;