Files
rustfs/crates/heal/tests/heal_integration_test.rs
T
houseme 27468ebfa9 feat(get): consolidate GET performance optimization (#3972)
* feat(get): consolidate GET performance optimization

Consolidated implementation of all GET performance optimizations into
a single, well-organized commit replacing the previous patch-on-patch
approach.

## Changes

### Configuration (set_disk/mod.rs)
- Consolidated all GET optimization flags into a single organized section
- Enabled by default: codec streaming, metadata early-stop, page cache reclaim
- Added codec streaming multipart flag (default: disabled)
- Added version-aware early-stop flag (default: disabled)
- Added adaptive duplex buffer sizing based on object size
- All flags use OnceLock caching with rollout percentage support

### Metadata Early-Stop (set_disk/read.rs)
- Delete marker early-stop when quorum agrees
- Version-aware early-stop for versioned GET requests
- MetadataQuorumAccumulator enhanced with:
  - delete_marker_votes tracking
  - requested_version_id and matching_version_votes tracking
  - version_early_stop_decision() method
- 6 new tests for version early-stop scenarios

### Codec Streaming (erasure/coding/decode_reader.rs)
- DualInFlight (2-stripe lookahead) enabled by default

### Decode Pipeline (erasure/coding/decode.rs)
- Stripe prefetch count configuration
- Bitrot-decode overlap configuration

### Disk Layer (disk/local.rs)
- O_DIRECT read configuration constants (preparation)

### Metrics (io-metrics/lib.rs)
- BytesPool acquisition/return metrics
- Metadata phase duration with early-stop label
- Total duration with reader_path label

### Diagnostics (diagnostics/)
- Early-stop reason constants
- Pool tier/outcome label constants

### Observability (.docker/observability/)
- 3 Grafana dashboards for GET optimization monitoring
- Prometheus alert rules (6 alerts: 3 critical, 3 warning)
- Updated README.md and README_ZH.md with usage docs

### Config (config/src/constants/runtime.rs)
- Page cache reclaim read enabled by default

## Environment Variables

| Variable | Default | Description |
|----------|---------|-------------|
| RUSTFS_GET_CODEC_STREAMING_ENABLE | true | Codec streaming base flag |
| RUSTFS_GET_CODEC_STREAMING_ROLLOUT_PCT | 100 | Codec streaming rollout % |
| RUSTFS_GET_CODEC_STREAMING_MULTIPART_ENABLE | false | Multipart codec streaming |
| RUSTFS_GET_METADATA_EARLY_STOP_ENABLE | true | Early-stop base flag |
| RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT | 100 | Early-stop rollout % |
| RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE | false | Version-aware early-stop |
| RUSTFS_OBJECT_FILE_CACHE_RECLAIM_READ_ENABLE | true | Page cache reclaim |
| RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE | false | O_DIRECT (preparation) |
| RUSTFS_GET_DECODE_STRIPE_PREFETCH_COUNT | 1 | Stripe prefetch |
| RUSTFS_GET_BITROT_DECODE_OVERLAP_ENABLE | false | Bitrot-decode overlap |
| RUSTFS_GET_CODEC_STREAMING_MAX_INFLIGHT | 2 | DualInFlight stripes |

## Rollback

All optimizations can be disabled via environment variables:
RUSTFS_GET_CODEC_STREAMING_ENABLE=false
RUSTFS_GET_METADATA_EARLY_STOP_ENABLE=false
RUSTFS_OBJECT_FILE_CACHE_RECLAIM_READ_ENABLE=false

Co-Authored-By: heihutu <heihutu@gmail.com>

* test(get): add stress test scripts for GET optimization validation

- quick-validate-get-optimization.sh: Quick 5-minute validation
- stress-test-get-optimization.sh: Full 30+ minute stress test
- README-stress-test.md: Usage documentation

Co-Authored-By: heihutu <heihutu@gmail.com>

* test(ecstore): align file cache reclaim defaults

* chore(deps): update redis and erasure codec

* test(ecstore): align decode fill policy default

* test(ecstore): align metadata early-stop default

* fix(ecstore): keep metadata early stop opt-in

---------

Co-authored-by: heihutu <heihutu@gmail.com>
2026-06-28 07:14:07 +08:00

473 lines
18 KiB
Rust

// 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 http::HeaderMap;
use rustfs_common::heal_channel::{HealOpts, HealScanMode};
use rustfs_heal::heal::{
manager::{HealConfig, HealManager},
storage::{ECStoreHealStorage, HealObjectOptions as ObjectOptions, HealPutObjReader as PutObjReader, HealStorageAPI},
task::{HealOptions, HealPriority, HealRequest, HealTaskStatus, HealType},
};
use serial_test::serial;
use std::{
path::{Path, PathBuf},
sync::{Arc, Once},
time::Duration,
};
use tokio::fs;
use tokio_util::sync::CancellationToken;
use tracing::info;
use walkdir::WalkDir;
#[path = "heal_integration_test/storage_api.rs"]
mod storage_api;
use storage_api::integration::{
BucketOperations, BucketOptions, ECStore, Endpoint, EndpointServerPools, Endpoints, ObjectIO as _, ObjectOperations as _,
PoolEndpoints, init_bucket_metadata_sys, init_local_disks,
};
const HEAL_FORMAT_WAIT_TIMEOUT: Duration = Duration::from_secs(25);
const HEAL_FORMAT_WAIT_INTERVAL: Duration = Duration::from_millis(250);
const NON_INLINE_TEST_DATA_SIZE: usize = 256 * 1024 + 137;
fn non_inline_test_data() -> Vec<u8> {
(0..NON_INLINE_TEST_DATA_SIZE).map(|idx| (idx % 251) as u8).collect()
}
async fn wait_for_path_exists(path: &Path, timeout: Duration, interval: Duration) -> bool {
let deadline = tokio::time::Instant::now() + timeout;
loop {
if path.exists() {
return true;
}
if tokio::time::Instant::now() >= deadline {
return false;
}
tokio::time::sleep(interval).await;
}
}
static INIT: Once = Once::new();
pub fn init_tracing() {
INIT.call_once(|| {
let _ = tracing_subscriber::fmt()
.with_env_filter(tracing_subscriber::EnvFilter::from_default_env())
.with_timer(tracing_subscriber::fmt::time::UtcTime::rfc_3339())
.with_thread_names(true)
.try_init();
});
}
/// Test helper: Create test environment with ECStore
async fn setup_test_env() -> (Vec<PathBuf>, Arc<ECStore>, Arc<ECStoreHealStorage>) {
init_tracing();
// create temp dir as 4 disks with unique base dir
let test_base_dir = format!("/tmp/rustfs_heal_heal_test_{}", uuid::Uuid::new_v4());
let temp_dir = std::path::PathBuf::from(&test_base_dir);
if temp_dir.exists() {
fs::remove_dir_all(&temp_dir).await.ok();
}
fs::create_dir_all(&temp_dir).await.unwrap();
// create 4 disk dirs
let disk_paths = vec![
temp_dir.join("disk1"),
temp_dir.join("disk2"),
temp_dir.join("disk3"),
temp_dir.join("disk4"),
];
for disk_path in &disk_paths {
fs::create_dir_all(disk_path).await.unwrap();
}
// create EndpointServerPools
let mut endpoints = Vec::new();
for (i, disk_path) in disk_paths.iter().enumerate() {
let mut endpoint = Endpoint::try_from(disk_path.to_str().unwrap()).unwrap();
// set correct index
endpoint.set_pool_index(0);
endpoint.set_set_index(0);
endpoint.set_disk_index(i);
endpoints.push(endpoint);
}
let pool_endpoints = PoolEndpoints {
legacy: false,
set_count: 1,
drives_per_set: 4,
endpoints: Endpoints::from(endpoints),
cmd_line: "test".to_string(),
platform: format!("OS: {} | Arch: {}", std::env::consts::OS, std::env::consts::ARCH),
};
let endpoint_pools = EndpointServerPools::from(vec![pool_endpoints]);
// format disks (only first time)
init_local_disks(endpoint_pools.clone()).await.unwrap();
// Use port 0 so nextest can run this integration binary in parallel
// with other ECStore-backed tests without sharing a fixed peer port.
let server_addr: std::net::SocketAddr = "127.0.0.1:0".parse().unwrap();
let ecstore = ECStore::new(server_addr, endpoint_pools, CancellationToken::new())
.await
.unwrap();
// init bucket metadata system
let buckets_list = ecstore
.list_bucket(&BucketOptions {
no_metadata: true,
..Default::default()
})
.await
.unwrap();
let buckets = buckets_list.into_iter().map(|v| v.name).collect();
init_bucket_metadata_sys(ecstore.clone(), buckets).await;
// Create heal storage layer
let heal_storage = Arc::new(ECStoreHealStorage::new(ecstore.clone()));
(disk_paths, ecstore, heal_storage)
}
/// Test helper: Create a test bucket
async fn create_test_bucket(ecstore: &Arc<ECStore>, bucket_name: &str) {
(**ecstore)
.make_bucket(bucket_name, &Default::default())
.await
.expect("Failed to create test bucket");
info!("Created test bucket: {}", bucket_name);
}
/// Test helper: Upload test object
async fn upload_test_object(ecstore: &Arc<ECStore>, bucket: &str, object: &str, data: &[u8]) {
let mut reader = PutObjReader::from_vec(data.to_vec());
let object_info = (**ecstore)
.put_object(bucket, object, &mut reader, &ObjectOptions::default())
.await
.expect("Failed to upload test object");
info!("Uploaded test object: {}/{} ({} bytes)", bucket, object, object_info.size);
}
mod serial_tests {
use super::*;
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
#[serial]
async fn test_heal_object_basic() {
let (disk_paths, ecstore, heal_storage) = setup_test_env().await;
// Create test bucket and object
let bucket_name = "test-heal-object-basic";
let object_name = "test-object.txt";
let test_data = non_inline_test_data();
create_test_bucket(&ecstore, bucket_name).await;
upload_test_object(&ecstore, bucket_name, object_name, &test_data).await;
let _obj_dir = disk_paths[0].join(bucket_name).join(object_name);
// ─── 1️⃣ delete single data shard file ─────────────────────────────────────
let obj_dir = disk_paths[0].join(bucket_name).join(object_name);
// find part file at depth 2, e.g. .../<uuid>/part.1
let target_part = WalkDir::new(&obj_dir)
.min_depth(2)
.max_depth(2)
.into_iter()
.filter_map(Result::ok)
.find(|e| e.file_type().is_file() && e.file_name().to_str().map(|n| n.starts_with("part.")).unwrap_or(false))
.map(|e| e.into_path())
.expect("Failed to locate part file to delete");
std::fs::remove_file(&target_part).expect("failed to delete part file");
assert!(!target_part.exists());
println!("✅ Deleted shard part file: {target_part:?}");
let heal_opts = HealOpts {
recreate: true,
remove: false,
update_parity: true,
..Default::default()
};
let (object_result, object_error) = heal_storage
.heal_object(bucket_name, object_name, None, &heal_opts)
.await
.expect("failed to heal object");
info!("heal_object result: {:?}, error: {:?}", object_result, object_error);
assert!(object_error.is_none(), "heal_object returned error: {object_error:?}");
// `test_heal_format_with_data` covers on-disk shard restoration. Here we
// focus on the object-level healing contract: the object must remain
// readable with intact contents after healing.
let mut reader = ecstore
.get_object_reader(bucket_name, object_name, None, HeaderMap::new(), &ObjectOptions::default())
.await
.expect("Failed to get object reader after heal");
let mut downloaded_data = Vec::new();
tokio::io::copy(&mut reader, &mut downloaded_data)
.await
.expect("Failed to read healed object data");
assert_eq!(downloaded_data, test_data, "Healed object data does not match original");
info!("Heal object basic test passed");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
#[serial]
async fn test_heal_bucket_basic() {
let (disk_paths, ecstore, heal_storage) = setup_test_env().await;
// Create test bucket
let bucket_name = "test-heal-bucket-basic";
create_test_bucket(&ecstore, bucket_name).await;
// ─── 1️⃣ delete bucket dir on disk ──────────────
let broken_bucket_path = disk_paths[0].join(bucket_name);
assert!(broken_bucket_path.exists(), "bucket dir does not exist on disk");
std::fs::remove_dir_all(&broken_bucket_path).expect("failed to delete bucket dir on disk");
assert!(!broken_bucket_path.exists(), "bucket dir still exists after deletion");
println!("✅ Deleted bucket directory on disk: {broken_bucket_path:?}");
// Create heal manager with faster interval
let cfg = HealConfig {
heal_interval: Duration::from_millis(1),
..Default::default()
};
let heal_manager = HealManager::new(heal_storage.clone(), Some(cfg));
heal_manager.start().await.unwrap();
// Submit heal request for the bucket
let heal_request = HealRequest::new(
HealType::Bucket {
bucket: bucket_name.to_string(),
},
HealOptions {
dry_run: false,
recursive: true,
remove_corrupted: false,
recreate_missing: false,
scan_mode: HealScanMode::Normal,
update_parity: false,
timeout: Some(Duration::from_secs(300)),
pool_index: None,
set_index: None,
},
HealPriority::Normal,
);
let task_id = heal_request.id.clone();
let admission = heal_manager
.submit_heal_request(heal_request)
.await
.expect("Failed to submit bucket heal request");
assert!(admission.is_admitted(), "bucket heal request should be admitted");
info!("Submitted bucket heal request with task ID: {}", task_id);
// Wait for task completion
tokio::time::sleep(tokio::time::Duration::from_secs(5)).await;
// Attempt to fetch task status (optional)
if let Ok(status) = heal_manager.get_task_status(&task_id).await {
if status == HealTaskStatus::Completed {
info!("Bucket heal task status: {:?}", status);
} else {
panic!("Bucket heal task status: {status:?}");
}
}
// ─── 3️⃣ Verify bucket directory is restored on every disk ───────
assert!(broken_bucket_path.exists(), "bucket dir does not exist on disk");
info!("Heal bucket basic test passed");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
#[serial]
async fn test_heal_format_basic() {
let (disk_paths, _ecstore, heal_storage) = setup_test_env().await;
// ─── 1️⃣ delete format.json on one disk ──────────────
let format_path = disk_paths[0].join(".rustfs.sys").join("format.json");
assert!(format_path.exists(), "format.json does not exist on disk");
std::fs::remove_file(&format_path).expect("failed to delete format.json on disk");
assert!(!format_path.exists(), "format.json still exists after deletion");
println!("✅ Deleted format.json on disk: {format_path:?}");
let (_format_result, format_error) = heal_storage.heal_format(false).await.expect("failed to run heal_format");
if let Some(err) = format_error {
info!("heal_format returned error: {:?}", err);
}
let restored = wait_for_path_exists(&format_path, HEAL_FORMAT_WAIT_TIMEOUT, HEAL_FORMAT_WAIT_INTERVAL).await;
assert!(restored, "format.json does not exist on disk after heal");
// ─── 2️⃣ verify format.json is restored ───────
assert!(format_path.exists(), "format.json does not exist on disk after heal");
info!("Heal format basic test passed");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
#[serial]
async fn test_heal_format_with_data() {
let (disk_paths, ecstore, heal_storage) = setup_test_env().await;
// Create test bucket and object
let bucket_name = "test-heal-format-with-data";
let object_name = "test-object.txt";
let test_data = non_inline_test_data();
create_test_bucket(&ecstore, bucket_name).await;
upload_test_object(&ecstore, bucket_name, object_name, &test_data).await;
let obj_dir = disk_paths[0].join(bucket_name).join(object_name);
let target_part = WalkDir::new(&obj_dir)
.min_depth(2)
.max_depth(2)
.into_iter()
.filter_map(Result::ok)
.find(|e| e.file_type().is_file() && e.file_name().to_str().map(|n| n.starts_with("part.")).unwrap_or(false))
.map(|e| e.into_path())
.expect("Failed to locate part file to delete");
// ─── 1️⃣ delete format.json on one disk ──────────────
let format_path = disk_paths[0].join(".rustfs.sys").join("format.json");
std::fs::remove_dir_all(&disk_paths[0]).expect("failed to delete all contents under disk_paths[0]");
std::fs::create_dir_all(&disk_paths[0]).expect("failed to recreate disk_paths[0] directory");
println!("✅ Deleted format.json on disk: {:?}", disk_paths[0]);
let (_format_result, format_error) = heal_storage.heal_format(false).await.expect("failed to run heal_format");
if let Some(err) = format_error {
info!("heal_format returned warning/error: {:?}", err);
}
let bucket_heal_opts = HealOpts {
recursive: true,
recreate: true,
..Default::default()
};
heal_storage
.heal_bucket(bucket_name, &bucket_heal_opts)
.await
.expect("failed to heal bucket");
let heal_opts = HealOpts {
recreate: true,
remove: false,
..Default::default()
};
let (object_result, object_error) = heal_storage
.heal_object(bucket_name, object_name, None, &heal_opts)
.await
.expect("failed to heal object");
info!("heal_object result: {:?}, error: {:?}", object_result, object_error);
assert!(object_error.is_none(), "heal_object returned error: {object_error:?}");
let format_restored = wait_for_path_exists(&format_path, HEAL_FORMAT_WAIT_TIMEOUT, HEAL_FORMAT_WAIT_INTERVAL).await;
assert!(format_restored, "format.json does not exist on disk after heal");
let target_restored = wait_for_path_exists(&target_part, HEAL_FORMAT_WAIT_TIMEOUT, HEAL_FORMAT_WAIT_INTERVAL).await;
// ─── 3️⃣ verify format.json is restored ───────
assert!(format_path.exists(), "format.json does not exist on disk after heal");
assert!(target_restored, "part file was not restored after heal");
// ─── 3️⃣ verify each part file is restored ───────
assert!(target_part.exists());
// Verify object metadata is accessible
let obj_info = ecstore
.get_object_info(bucket_name, object_name, &ObjectOptions::default())
.await
.expect("Expected object to be readable after heal");
assert_eq!(obj_info.size as usize, test_data.len());
// Actually read the object data to verify integrity
let mut reader = ecstore
.get_object_reader(bucket_name, object_name, None, HeaderMap::new(), &ObjectOptions::default())
.await
.expect("Failed to get object reader");
let mut downloaded_data = Vec::new();
tokio::io::copy(&mut reader, &mut downloaded_data)
.await
.expect("Failed to read object data");
assert_eq!(downloaded_data, test_data, "Healed object data does not match original");
info!("Heal format with data test passed");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
#[serial]
async fn test_heal_storage_api_direct() {
let (_disk_paths, ecstore, heal_storage) = setup_test_env().await;
// Test direct heal storage API calls
// Test heal_format
let format_result = heal_storage.heal_format(true).await; // dry run
assert!(format_result.is_ok());
info!("Direct heal_format test passed");
// Test heal_bucket
let bucket_name = "test-bucket-direct";
create_test_bucket(&ecstore, bucket_name).await;
let heal_opts = HealOpts {
recursive: true,
dry_run: true,
remove: false,
recreate: false,
scan_mode: HealScanMode::Normal,
update_parity: false,
no_lock: false,
pool: None,
set: None,
};
let bucket_result = heal_storage.heal_bucket(bucket_name, &heal_opts).await;
assert!(bucket_result.is_ok());
info!("Direct heal_bucket test passed");
// Test heal_object
let object_name = "test-object-direct.txt";
let test_data = b"Test data for direct heal API";
upload_test_object(&ecstore, bucket_name, object_name, test_data).await;
let object_heal_opts = HealOpts {
recursive: false,
dry_run: true,
remove: false,
recreate: false,
scan_mode: HealScanMode::Normal,
update_parity: false,
no_lock: false,
pool: None,
set: None,
};
let object_result = heal_storage
.heal_object(bucket_name, object_name, None, &object_heal_opts)
.await;
assert!(object_result.is_ok());
info!("Direct heal_object test passed");
info!("Direct heal storage API test passed");
}
}