mirror of
https://github.com/rustfs/rustfs.git
synced 2026-08-19 02:56:18 +00:00
fix(storage): harden offline drive fail-fast paths (#2564)
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: houseme <4829346+houseme@users.noreply.github.com>
This commit is contained in:
@@ -0,0 +1,60 @@
|
||||
// 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.
|
||||
|
||||
/// Legacy global drive timeout fallback.
|
||||
/// Deprecated in favor of per-operation drive timeout knobs.
|
||||
pub const ENV_DRIVE_MAX_TIMEOUT_DURATION: &str = "RUSTFS_DRIVE_MAX_TIMEOUT_DURATION";
|
||||
|
||||
/// Default timeout in seconds for the legacy global drive timeout fallback.
|
||||
pub const DEFAULT_DRIVE_MAX_TIMEOUT_DURATION_SECS: u64 = 30;
|
||||
|
||||
/// Timeout for metadata-oriented drive operations such as `read_metadata`.
|
||||
pub const ENV_DRIVE_METADATA_TIMEOUT_SECS: &str = "RUSTFS_DRIVE_METADATA_TIMEOUT_SECS";
|
||||
pub const DEFAULT_DRIVE_METADATA_TIMEOUT_SECS: u64 = 5;
|
||||
|
||||
/// Timeout for `disk_info()` calls on local and remote drives.
|
||||
pub const ENV_DRIVE_DISK_INFO_TIMEOUT_SECS: &str = "RUSTFS_DRIVE_DISK_INFO_TIMEOUT_SECS";
|
||||
pub const DEFAULT_DRIVE_DISK_INFO_TIMEOUT_SECS: u64 = 5;
|
||||
|
||||
/// Timeout for `list_dir()` style metadata listing operations.
|
||||
pub const ENV_DRIVE_LIST_DIR_TIMEOUT_SECS: &str = "RUSTFS_DRIVE_LIST_DIR_TIMEOUT_SECS";
|
||||
pub const DEFAULT_DRIVE_LIST_DIR_TIMEOUT_SECS: u64 = 5;
|
||||
|
||||
/// Total timeout for `walk_dir()` operations.
|
||||
pub const ENV_DRIVE_WALKDIR_TIMEOUT_SECS: &str = "RUSTFS_DRIVE_WALKDIR_TIMEOUT_SECS";
|
||||
pub const DEFAULT_DRIVE_WALKDIR_TIMEOUT_SECS: u64 = 5;
|
||||
|
||||
/// Maximum time without forward progress while consuming a `walk_dir()` stream.
|
||||
pub const ENV_DRIVE_WALKDIR_STALL_TIMEOUT_SECS: &str = "RUSTFS_DRIVE_WALKDIR_STALL_TIMEOUT_SECS";
|
||||
pub const DEFAULT_DRIVE_WALKDIR_STALL_TIMEOUT_SECS: u64 = 5;
|
||||
|
||||
/// Number of consecutive failures before a suspect drive is classified as offline.
|
||||
pub const ENV_DRIVE_SUSPECT_FAILURE_THRESHOLD: &str = "RUSTFS_DRIVE_SUSPECT_FAILURE_THRESHOLD";
|
||||
pub const DEFAULT_DRIVE_SUSPECT_FAILURE_THRESHOLD: u64 = 2;
|
||||
|
||||
/// Number of consecutive successful recovery probes before a returning drive is considered online again.
|
||||
pub const ENV_DRIVE_RETURNING_SUCCESS_THRESHOLD: &str = "RUSTFS_DRIVE_RETURNING_SUCCESS_THRESHOLD";
|
||||
pub const DEFAULT_DRIVE_RETURNING_SUCCESS_THRESHOLD: u64 = 3;
|
||||
|
||||
/// Probe interval in seconds while a drive is in the recovery path.
|
||||
pub const ENV_DRIVE_RETURNING_PROBE_INTERVAL_SECS: &str = "RUSTFS_DRIVE_RETURNING_PROBE_INTERVAL_SECS";
|
||||
pub const DEFAULT_DRIVE_RETURNING_PROBE_INTERVAL_SECS: u64 = 2;
|
||||
|
||||
/// Duration in seconds for classifying a recovered drive as a short offline event.
|
||||
pub const ENV_DRIVE_OFFLINE_GRACE_PERIOD_SECS: &str = "RUSTFS_DRIVE_OFFLINE_GRACE_PERIOD_SECS";
|
||||
pub const DEFAULT_DRIVE_OFFLINE_GRACE_PERIOD_SECS: u64 = 30;
|
||||
|
||||
/// Duration in seconds after which a recovered drive is classified as long offline.
|
||||
pub const ENV_DRIVE_LONG_OFFLINE_THRESHOLD_SECS: &str = "RUSTFS_DRIVE_LONG_OFFLINE_THRESHOLD_SECS";
|
||||
pub const DEFAULT_DRIVE_LONG_OFFLINE_THRESHOLD_SECS: u64 = 172_800;
|
||||
@@ -56,6 +56,15 @@ pub const ENV_HEAL_TASK_TIMEOUT_SECS: &str = "RUSTFS_HEAL_TASK_TIMEOUT_SECS";
|
||||
/// - Note: A higher concurrency limit can speed up healing but may lead to resource contention.
|
||||
pub const ENV_HEAL_MAX_CONCURRENT_HEALS: &str = "RUSTFS_HEAL_MAX_CONCURRENT_HEALS";
|
||||
|
||||
/// Environment variable name that specifies the maximum number of concurrent heal operations
|
||||
/// allowed for a single erasure set.
|
||||
///
|
||||
/// - Purpose: Prevent one degraded set from consuming all global heal slots.
|
||||
/// - Unit: number of operations (usize).
|
||||
/// - Valid values: any positive integer.
|
||||
/// - Example: `export RUSTFS_HEAL_MAX_CONCURRENT_PER_SET=1`
|
||||
pub const ENV_HEAL_MAX_CONCURRENT_PER_SET: &str = "RUSTFS_HEAL_MAX_CONCURRENT_PER_SET";
|
||||
|
||||
/// Default value for enabling authentication for heal operations if not specified in the environment variable.
|
||||
/// - Value: true (authentication enabled).
|
||||
/// - Rationale: Enabling authentication by default enhances security for heal operations.
|
||||
@@ -86,3 +95,46 @@ pub const DEFAULT_HEAL_TASK_TIMEOUT_SECS: u64 = 300; // 5 minutes
|
||||
/// - Rationale: This default concurrency limit helps balance healing speed with resource usage, preventing system overload.
|
||||
/// - Adjustments: Users may modify this value via the `RUSTFS_HEAL_MAX_CONCURRENT_HEALS` environment variable based on their system capacity and expected heal workload.
|
||||
pub const DEFAULT_HEAL_MAX_CONCURRENT_HEALS: usize = 4;
|
||||
|
||||
/// Default maximum number of concurrent heal operations per erasure set.
|
||||
///
|
||||
/// - Value: 1 concurrent heal operation per set.
|
||||
/// - Rationale: Keeps a degraded set from monopolizing the global heal scheduler.
|
||||
pub const DEFAULT_HEAL_MAX_CONCURRENT_PER_SET: usize = 1;
|
||||
|
||||
/// Environment variable that controls whether low-priority heal requests should merge into
|
||||
/// an existing queued request with the same deduplication key.
|
||||
pub const ENV_HEAL_LOW_PRIORITY_MERGE_ENABLE: &str = "RUSTFS_HEAL_LOW_PRIORITY_MERGE_ENABLE";
|
||||
|
||||
/// Environment variable that allows low-priority heal requests to be dropped when the queue is full.
|
||||
pub const ENV_HEAL_LOW_PRIORITY_DROP_WHEN_FULL: &str = "RUSTFS_HEAL_LOW_PRIORITY_DROP_WHEN_FULL";
|
||||
|
||||
/// Environment variable that controls concurrent object heals within a single erasure-set page.
|
||||
pub const ENV_HEAL_PAGE_OBJECT_CONCURRENCY: &str = "RUSTFS_HEAL_PAGE_OBJECT_CONCURRENCY";
|
||||
|
||||
/// Environment variable that toggles notify-driven scheduler wakeups.
|
||||
pub const ENV_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE: &str = "RUSTFS_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE";
|
||||
|
||||
/// Environment variable that toggles per-set bulkhead scheduling.
|
||||
pub const ENV_HEAL_SET_BULKHEAD_ENABLE: &str = "RUSTFS_HEAL_SET_BULKHEAD_ENABLE";
|
||||
|
||||
/// Environment variable that toggles page-level parallel object healing for erasure-set repair.
|
||||
pub const ENV_HEAL_PAGE_PARALLEL_ENABLE: &str = "RUSTFS_HEAL_PAGE_PARALLEL_ENABLE";
|
||||
|
||||
/// Default behavior is to merge duplicate low-priority requests.
|
||||
pub const DEFAULT_HEAL_LOW_PRIORITY_MERGE_ENABLE: bool = true;
|
||||
|
||||
/// Default behavior is to drop low-priority requests instead of blocking when the queue is full.
|
||||
pub const DEFAULT_HEAL_LOW_PRIORITY_DROP_WHEN_FULL: bool = true;
|
||||
|
||||
/// Default per-page object heal concurrency for erasure-set healing.
|
||||
pub const DEFAULT_HEAL_PAGE_OBJECT_CONCURRENCY: usize = 8;
|
||||
|
||||
/// Default behavior is to keep notify-driven scheduler wakeups enabled.
|
||||
pub const DEFAULT_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE: bool = true;
|
||||
|
||||
/// Default behavior is to keep per-set bulkhead scheduling enabled.
|
||||
pub const DEFAULT_HEAL_SET_BULKHEAD_ENABLE: bool = true;
|
||||
|
||||
/// Default behavior is to keep erasure-set page parallelism enabled.
|
||||
pub const DEFAULT_HEAL_PAGE_PARALLEL_ENABLE: bool = true;
|
||||
|
||||
@@ -17,6 +17,7 @@ pub(crate) mod body_limits;
|
||||
pub(crate) mod capacity;
|
||||
pub(crate) mod compress;
|
||||
pub(crate) mod console;
|
||||
pub(crate) mod drive;
|
||||
pub(crate) mod env;
|
||||
pub(crate) mod heal;
|
||||
pub(crate) mod object;
|
||||
|
||||
@@ -42,6 +42,15 @@ pub const ENV_SCANNER_IDLE_MODE: &str = "RUSTFS_SCANNER_IDLE_MODE";
|
||||
/// Default scanner idle mode.
|
||||
pub const DEFAULT_SCANNER_IDLE_MODE: bool = true;
|
||||
|
||||
/// Compatibility flag kept for Patch 3 rollback windows.
|
||||
///
|
||||
/// Inline scanner heal execution has been removed in favor of heal-candidate enqueue.
|
||||
/// When this flag is enabled, RustFS logs a warning and continues to use enqueue-based heal.
|
||||
pub const ENV_SCANNER_INLINE_HEAL_ENABLE: &str = "RUSTFS_SCANNER_INLINE_HEAL_ENABLE";
|
||||
|
||||
/// Default inline scanner heal compatibility mode.
|
||||
pub const DEFAULT_SCANNER_INLINE_HEAL_ENABLE: bool = false;
|
||||
|
||||
/// Scanner speed preset controlling throttling behavior.
|
||||
///
|
||||
/// Each preset defines three parameters:
|
||||
|
||||
@@ -25,6 +25,8 @@ pub use constants::compress::*;
|
||||
#[cfg(feature = "constants")]
|
||||
pub use constants::console::*;
|
||||
#[cfg(feature = "constants")]
|
||||
pub use constants::drive::*;
|
||||
#[cfg(feature = "constants")]
|
||||
pub use constants::env::*;
|
||||
#[cfg(feature = "constants")]
|
||||
pub use constants::heal::*;
|
||||
|
||||
Reference in New Issue
Block a user