mirror of
https://github.com/rustfs/rustfs.git
synced 2026-08-01 02:52:15 +00:00
Compare commits
2 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
| ba981b9d83 | |||
| 5142c7da27 |
@@ -300,9 +300,6 @@ path = "junit.xml"
|
||||
# negative-path siblings of each family stay in as regression guards.
|
||||
# * rustfs#4843 — over-limit archive entry paths hard-reject the whole
|
||||
# archive even under ignore-errors semantics.
|
||||
# * rustfs#4846 — distributed-lock quorum tests misclassify as timeout
|
||||
# under parallel load (multi-node in-process clusters; natural home is
|
||||
# ci-7's nightly cluster lane).
|
||||
[profile.e2e-full]
|
||||
default-filter = """
|
||||
package(e2e_test)
|
||||
@@ -311,7 +308,6 @@ default-filter = """
|
||||
& !test(/^replication_extension_test::/)
|
||||
& !test(/^multipart_auth_test::test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_enabled$/)
|
||||
& !test(/^snowball_auto_extract_test::tests::snowball_auto_extract_(ignores_invalid_entries_when_requested|supports_standard_headers_with_combined_extract_options)$/)
|
||||
& !test(/^reliant::lock::test_distributed_lock_(2_nodes_grpc_read_survives_failed_node|4_nodes_grpc_read_write_quorum_split_with_two_failed_nodes)$/)
|
||||
"""
|
||||
fail-fast = false
|
||||
|
||||
|
||||
@@ -15,9 +15,7 @@
|
||||
|
||||
use super::{grpc_lock_client::GrpcLockClient, grpc_lock_server::spawn_lock_server};
|
||||
use rustfs_lock::client::{LockClient, local::LocalClient};
|
||||
use rustfs_lock::{
|
||||
GlobalLockManager, LockError, LockInfo, LockRequest, LockResponse, LockStats, LockType, NamespaceLock, ObjectKey,
|
||||
};
|
||||
use rustfs_lock::{GlobalLockManager, LockInfo, LockRequest, LockResponse, LockStats, LockType, NamespaceLock, ObjectKey};
|
||||
use std::sync::Arc;
|
||||
use std::time::Duration;
|
||||
|
||||
@@ -35,7 +33,11 @@ struct FailingClient;
|
||||
#[async_trait::async_trait]
|
||||
impl rustfs_lock::LockClient for FailingClient {
|
||||
async fn acquire_lock(&self, _request: &rustfs_lock::LockRequest) -> rustfs_lock::Result<LockResponse> {
|
||||
Err(LockError::internal("simulated gRPC node failure"))
|
||||
// Match RemoteClient's transport-failure response so the coordinator can count this node toward quorum loss.
|
||||
Ok(LockResponse::failure(
|
||||
"Remote lock RPC failed: simulated gRPC node failure",
|
||||
Duration::ZERO,
|
||||
))
|
||||
}
|
||||
|
||||
async fn release(&self, _lock_id: &rustfs_lock::LockId) -> rustfs_lock::Result<bool> {
|
||||
|
||||
@@ -382,6 +382,7 @@ struct ManualTransitionRunReport {
|
||||
skipped_delete_marker: u64,
|
||||
skipped_directory: u64,
|
||||
skipped_replication: u64,
|
||||
skipped_already_transitioned: u64,
|
||||
skipped_already_in_flight: u64,
|
||||
skipped_queue_full: u64,
|
||||
skipped_queue_closed: u64,
|
||||
@@ -407,6 +408,41 @@ fn assert_completed_or_in_flight_partial(state: &str, report: &ManualTransitionR
|
||||
}
|
||||
}
|
||||
|
||||
fn assert_conflict_winner_report(state: &str, report: &ManualTransitionRunReport, expected_objects: u64, context: &str) {
|
||||
assert_completed_or_in_flight_partial(state, report, context);
|
||||
if report.skipped_already_in_flight > 0 {
|
||||
assert!(
|
||||
report.scanned <= expected_objects,
|
||||
"{context}: scanned more objects than the conflict scope contains: {report:#?}"
|
||||
);
|
||||
assert!(
|
||||
report.eligible <= expected_objects,
|
||||
"{context}: marked more objects eligible than the conflict scope contains: {report:#?}"
|
||||
);
|
||||
assert_eq!(
|
||||
report.enqueued + report.skipped_already_in_flight,
|
||||
report.eligible,
|
||||
"{context}: partial in-flight accounting must cover every eligible object: {report:#?}"
|
||||
);
|
||||
} else {
|
||||
assert_eq!(report.scanned, expected_objects, "{context}: {report:#?}");
|
||||
assert_eq!(
|
||||
report.eligible + report.skipped_already_transitioned,
|
||||
expected_objects,
|
||||
"{context}: {report:#?}"
|
||||
);
|
||||
assert_eq!(
|
||||
report.enqueued + report.skipped_already_in_flight,
|
||||
expected_objects,
|
||||
"{context}: {report:#?}"
|
||||
);
|
||||
}
|
||||
assert_eq!(
|
||||
report.transition_completed, report.enqueued,
|
||||
"{context}: winner must wait for all queued transitions: {report:#?}"
|
||||
);
|
||||
}
|
||||
|
||||
#[derive(Debug, Deserialize)]
|
||||
struct ManualTransitionQueueSnapshot {
|
||||
queue_capacity: u64,
|
||||
@@ -1240,15 +1276,8 @@ async fn test_manual_transition_async_scope_conflicts_report_active_job() -> Tes
|
||||
cold_client.create_bucket().bucket(TIER_BUCKET).send().await?;
|
||||
|
||||
let mut hot = RustFSTestEnvironment::new().await?;
|
||||
hot.start_rustfs_server_with_env(
|
||||
vec![],
|
||||
&[
|
||||
("RUSTFS_SCANNER_ENABLED", "false"),
|
||||
("RUSTFS_SCANNER_CYCLE", "3600"),
|
||||
(MANUAL_TRANSITION_CANCEL_BARRIER_ENV, "1"),
|
||||
],
|
||||
)
|
||||
.await?;
|
||||
hot.start_rustfs_server_with_env(vec![], &[("RUSTFS_SCANNER_ENABLED", "false"), ("RUSTFS_SCANNER_CYCLE", "3600")])
|
||||
.await?;
|
||||
let hot_client = hot.create_s3_client();
|
||||
add_rustfs_tier(&hot, &cold).await?;
|
||||
|
||||
@@ -1318,21 +1347,20 @@ async fn test_manual_transition_async_scope_conflicts_report_active_job() -> Tes
|
||||
assert_eq!(conflict.cancel_endpoint, status_endpoint);
|
||||
assert!(!conflict.scope_key.is_empty());
|
||||
|
||||
manual_transition_job_cancel(&hot, cancel_endpoint).await?;
|
||||
|
||||
let terminal = wait_for_manual_transition_job_terminal(&hot, status_endpoint, MANUAL_ASYNC_CONFLICT_TERMINAL_TIMEOUT).await?;
|
||||
assert_eq!(terminal.job_id, job_id);
|
||||
assert_eq!(terminal.status, "cancelled", "terminal conflict winner response: {terminal:#?}");
|
||||
assert!(!terminal.report.dry_run);
|
||||
assert_eq!(terminal.report.bucket, MANUAL_ASYNC_CONFLICT_BUCKET);
|
||||
assert_eq!(terminal.report.prefix, accepted.report.prefix);
|
||||
assert!(terminal.report.cancelled, "terminal conflict winner response: {terminal:#?}");
|
||||
assert_eq!(terminal.report.scanned, 0, "terminal conflict winner response: {terminal:#?}");
|
||||
assert_eq!(terminal.report.enqueued, 0, "terminal conflict winner response: {terminal:#?}");
|
||||
assert_eq!(
|
||||
terminal.report.transition_completed, 0,
|
||||
"terminal conflict winner response: {terminal:#?}"
|
||||
assert_conflict_winner_report(
|
||||
&terminal.status,
|
||||
&terminal.report,
|
||||
MANUAL_ASYNC_CONFLICT_OBJECTS as u64,
|
||||
"terminal conflict winner response",
|
||||
);
|
||||
assert_eq!(terminal.report.dry_run_eligible, 0, "terminal conflict winner response: {terminal:#?}");
|
||||
assert_eq!(terminal.report.transition_failed, 0, "terminal conflict winner response: {terminal:#?}");
|
||||
assert_eq!(terminal.report.tier_failure, 0, "terminal conflict winner response: {terminal:#?}");
|
||||
let after_remote_count = cold_tier_object_count(&cold_client).await?;
|
||||
assert!(after_remote_count >= before_remote_count);
|
||||
assert!(after_remote_count <= before_remote_count + MANUAL_ASYNC_CONFLICT_OBJECTS);
|
||||
|
||||
@@ -75,7 +75,7 @@ const DATA_SCANNER_COMPACT_LEAST_OBJECT: usize = 500;
|
||||
const DATA_SCANNER_COMPACT_AT_CHILDREN: usize = 10000;
|
||||
const DATA_SCANNER_COMPACT_AT_FOLDERS: usize = DATA_SCANNER_COMPACT_AT_CHILDREN / 4;
|
||||
const DATA_SCANNER_FORCE_COMPACT_AT_FOLDERS: usize = 250_000;
|
||||
const SCANNER_LIST_PATH_RAW_STALL_TIMEOUT: Duration = Duration::from_secs(60);
|
||||
const SCANNER_LIST_PATH_RAW_TIMEOUT: Duration = Duration::from_secs(60);
|
||||
const SCANNER_ENTRY_PROGRESS_BATCH: u64 = 32;
|
||||
const SCANNER_ENTRY_PROGRESS_INTERVAL: Duration = Duration::from_secs(30);
|
||||
const DEFAULT_HEAL_OBJECT_SELECT_PROB: u32 = 1024;
|
||||
@@ -100,21 +100,6 @@ static SCANNER_INLINE_HEAL_WARN_ONCE: Once = Once::new();
|
||||
static SCANNER_INLINE_HEAL_METRICS_ONCE: Once = Once::new();
|
||||
static SCANNER_ALERT_METRICS_ONCE: Once = Once::new();
|
||||
|
||||
#[cfg(test)]
|
||||
type ListPathRawTimeoutSnapshot = (bool, Option<Duration>, Option<Duration>);
|
||||
|
||||
fn scanner_abandoned_child_list_options() -> ListPathRawOptions {
|
||||
// A complete heal walk scales with bucket size and may legitimately take
|
||||
// longer than a fixed wall-clock budget. Keep the total duration unbounded;
|
||||
// Retain the scanner's per-read stall budget and keep cancellation controlled
|
||||
// by the scanner cycle token.
|
||||
ListPathRawOptions {
|
||||
skip_walkdir_total_timeout: true,
|
||||
walkdir_stall_timeout: Some(SCANNER_LIST_PATH_RAW_STALL_TIMEOUT),
|
||||
..Default::default()
|
||||
}
|
||||
}
|
||||
|
||||
pub fn data_usage_update_dir_cycles() -> u32 {
|
||||
rustfs_utils::get_env_u32(ENV_DATA_USAGE_UPDATE_DIR_CYCLES, DATA_USAGE_UPDATE_DIR_CYCLES)
|
||||
}
|
||||
@@ -1296,8 +1281,6 @@ pub struct FolderScanner {
|
||||
skip_heal: Arc<std::sync::atomic::AtomicBool>,
|
||||
local_disk: Arc<Disk>,
|
||||
pending_heals_changed: bool,
|
||||
#[cfg(test)]
|
||||
list_path_raw_options_observer: Option<mpsc::UnboundedSender<ListPathRawTimeoutSnapshot>>,
|
||||
}
|
||||
|
||||
impl FolderScanner {
|
||||
@@ -2427,79 +2410,75 @@ impl FolderScanner {
|
||||
let bucket_clone = bucket.clone();
|
||||
let prefix_clone = prefix.clone();
|
||||
let child_ctx_clone = child_ctx.clone();
|
||||
#[cfg(test)]
|
||||
let list_path_raw_options_observer = self.list_path_raw_options_observer.clone();
|
||||
|
||||
tokio::spawn(async move {
|
||||
let options = ListPathRawOptions {
|
||||
disks,
|
||||
bucket: bucket_clone.clone(),
|
||||
path: prefix_clone.clone(),
|
||||
recursive: true,
|
||||
report_not_found: true,
|
||||
min_disks: disks_quorum,
|
||||
agreed: Some(Box::new(move |entry: MetaCacheEntry| {
|
||||
let entry_name = entry.name.clone();
|
||||
let agreed_tx = agreed_tx.clone();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = agreed_tx.send(entry_name).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
entry = %entry.name,
|
||||
state = "list_path_agreed_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw agreed callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
partial: Some(Box::new(move |entries: MetaCacheEntries, _: &[Option<DiskError>]| {
|
||||
let partial_tx = partial_tx.clone();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = partial_tx.send(entries).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
state = "list_path_partial_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw partial callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
finished: Some(Box::new(move |errs: &[Option<DiskError>]| {
|
||||
let finished_tx = finished_tx.clone();
|
||||
let errs_clone = errs.to_vec();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = finished_tx.send(errs_clone).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
state = "list_path_finished_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw finished callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
..scanner_abandoned_child_list_options()
|
||||
};
|
||||
#[cfg(test)]
|
||||
if let Some(observer) = list_path_raw_options_observer {
|
||||
let _ = observer.send((
|
||||
options.skip_walkdir_total_timeout,
|
||||
options.walkdir_timeout,
|
||||
options.walkdir_stall_timeout,
|
||||
));
|
||||
}
|
||||
if let Err(e) = list_path_raw(child_ctx_clone.clone(), options).await {
|
||||
if let Err(e) = list_path_raw(
|
||||
child_ctx_clone.clone(),
|
||||
ListPathRawOptions {
|
||||
disks,
|
||||
bucket: bucket_clone.clone(),
|
||||
path: prefix_clone.clone(),
|
||||
recursive: true,
|
||||
report_not_found: true,
|
||||
min_disks: disks_quorum,
|
||||
walkdir_timeout: Some(SCANNER_LIST_PATH_RAW_TIMEOUT),
|
||||
walkdir_stall_timeout: Some(SCANNER_LIST_PATH_RAW_TIMEOUT),
|
||||
agreed: Some(Box::new(move |entry: MetaCacheEntry| {
|
||||
let entry_name = entry.name.clone();
|
||||
let agreed_tx = agreed_tx.clone();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = agreed_tx.send(entry_name).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
entry = %entry.name,
|
||||
state = "list_path_agreed_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw agreed callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
partial: Some(Box::new(move |entries: MetaCacheEntries, _: &[Option<DiskError>]| {
|
||||
let partial_tx = partial_tx.clone();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = partial_tx.send(entries).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
state = "list_path_partial_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw partial callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
finished: Some(Box::new(move |errs: &[Option<DiskError>]| {
|
||||
let finished_tx = finished_tx.clone();
|
||||
let errs_clone = errs.to_vec();
|
||||
Box::pin(async move {
|
||||
if let Err(e) = finished_tx.send(errs_clone).await {
|
||||
error!(
|
||||
target: "rustfs::scanner::folder",
|
||||
event = EVENT_SCANNER_FOLDER_STATE,
|
||||
component = LOG_COMPONENT_SCANNER,
|
||||
subsystem = LOG_SUBSYSTEM_FOLDER,
|
||||
state = "list_path_finished_send_failed",
|
||||
error = %e,
|
||||
"Scanner list_path_raw finished callback failed"
|
||||
);
|
||||
}
|
||||
})
|
||||
})),
|
||||
..Default::default()
|
||||
},
|
||||
)
|
||||
.await
|
||||
{
|
||||
if is_missing_path_disk_error(&e) {
|
||||
debug!(
|
||||
target: "rustfs::scanner::folder",
|
||||
@@ -2841,8 +2820,6 @@ pub async fn scan_data_folder(
|
||||
skip_heal,
|
||||
local_disk,
|
||||
pending_heals_changed: false,
|
||||
#[cfg(test)]
|
||||
list_path_raw_options_observer: None,
|
||||
};
|
||||
|
||||
// Check if context is cancelled
|
||||
@@ -3049,7 +3026,6 @@ mod tests {
|
||||
skip_heal: Arc::new(AtomicBool::new(false)),
|
||||
local_disk: disk,
|
||||
pending_heals_changed: false,
|
||||
list_path_raw_options_observer: None,
|
||||
};
|
||||
|
||||
(scanner, temp_dir)
|
||||
@@ -4234,8 +4210,6 @@ mod tests {
|
||||
scanner.heal_object_select = 1;
|
||||
scanner.disks = disks;
|
||||
scanner.disks_quorum = 2;
|
||||
let (options_tx, mut options_rx) = mpsc::unbounded_channel();
|
||||
scanner.list_path_raw_options_observer = Some(options_tx);
|
||||
scanner.old_cache.replace(
|
||||
&format!("{bucket}/{object}"),
|
||||
bucket,
|
||||
@@ -4257,11 +4231,6 @@ mod tests {
|
||||
.expect("scan_folder should not hang after list_path_raw finishes")
|
||||
.expect("scan_folder should finish successfully");
|
||||
|
||||
let observed_options = tokio::time::timeout(Duration::from_secs(1), options_rx.recv())
|
||||
.await
|
||||
.expect("abandoned-child listing options should be observed promptly")
|
||||
.expect("abandoned-child listing options channel should remain open");
|
||||
assert_eq!(observed_options, (true, None, Some(SCANNER_LIST_PATH_RAW_STALL_TIMEOUT)));
|
||||
let root = scanner
|
||||
.new_cache
|
||||
.checked_flatten(bucket)
|
||||
|
||||
Reference in New Issue
Block a user