* fix(ecstore): skip hidden metadata in walk limit
* fix(ecstore): match walk limit to visible versions
* fix(ecstore): avoid decoding all versions for limit
* fix: return 503 on lock contention instead of 500
When concurrent PUTs to the same key contend for the namespace write
lock, lock timeout and conflict errors were wrapped as
StorageError::other(...) → StorageError::Io(...), which fell through
to S3ErrorCode::InternalError (500) in the error mapping.
Only QuorumNotReached was correctly mapped to ServiceUnavailable (503);
all other lock errors (Timeout, AlreadyLocked, etc.) became 500.
Fix: map_namespace_lock_error now returns StorageError::Lock(err) for
non-quorum lock errors, and StorageError::Lock is mapped to
S3ErrorCode::ServiceUnavailable in the HTTP error conversion.
Affected paths:
- crates/ecstore/src/set_disk/lock.rs
- crates/ecstore/src/store/object.rs
- crates/ecstore/src/store/bucket.rs (make_bucket, delete_bucket)
- rustfs/src/app/object_usecase.rs (copy_object)
Regression test: test_concurrent_put_same_key_never_returns_500
* test: fail on unexpected lock contention errors
---------
Co-authored-by: houseme <housemecn@gmail.com>
* fix(lifecycle): harden scanner ILM expiry accounting
* fix(scanner): gate ILM action accounting on enqueue
* fix(metrics): avoid scanner source work argument list
* fix(scanner): gate local ILM accounting on enqueue
---------
Co-authored-by: Henry Guo <marshawcoco@users.noreply.github.com>
Co-authored-by: 安正超 <anzhengchao@gmail.com>
Co-authored-by: houseme <housemecn@gmail.com>
Log local endpoint disk-topology details during physical disk independence checks.
Include canonical paths, st_dev major:minor values, and resolved device ids in validation output to make Docker and bind-mount startup failures easier to diagnose.
Also assert the new diagnostics are present in the shared-disk regression test.
* fix(tier): stop sending nil/garbage versionId to warm backend S3
Three bugs caused NoSuchVersion errors when reading tiered objects:
1. warm_backend_s3sdk: GET and DELETE ignored rv/range opts entirely —
fixed to forward version_id and byte-range to the SDK request.
2. version.rs (MetaObject + MetaDeleteMarker): transition_version_id was
parsed with unwrap_or_default(), turning invalid/wrong-length bytes
into Uuid::nil(). The nil UUID was then serialized and sent as
?versionId=00000000-... to the tier backend -> NoSuchVersion.
Fixed: .and_then(.ok()).filter(!is_nil()) so only valid non-nil UUIDs
are forwarded as versionId.
3. bucket_lifecycle_ops: add debug/error logs in
get_transitioned_object_reader to record tier, tier_object, and
tier_version_id before and on failure of the tier GET.
Also adds tier transition fields to dump_fileinfo example for offline
xl.meta inspection, and fixes Docker build (cargo path + entrypoint).
Adds CLAUDE.md with tier architecture and debugging notes.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* more fixes for versionId
* Potential fix for pull request finding
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Signed-off-by: Marcelo Bartsch <marcelo@bartsch.cl>
* remove branch
* Add tests and fix cargo path, add load to build-docker
* update documentation (CLAUDE.md)
* more fixes for recover
* More fixes to ILM recover
* final fix
* chore: add missing-shard first-scene diagnostics (#3213)
chore(ecstore): add missing-shard first-scene diagnostics
Log rename_data quorum context behind RUSTFS_ISSUE3031_DIAG_ENABLE so partial-disk success can be correlated with later missing shard reads.
Also log put_object commit success and tmp cleanup boundaries to capture when successful quorum writes are followed by tmp_dir cleanup.
* fix test anmd fmt
* fix cargo path
fix test
* fix(tier): format copy_object self-copy guard
---------
Signed-off-by: Marcelo Bartsch <marcelo@bartsch.cl>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: 安正超 <anzhengchao@gmail.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: houseme <housemecn@gmail.com>
Co-authored-by: cxymds <Cxymds@qq.com>
Co-authored-by: loverustfs <hello@rustfs.com>
* refactor(ecstore): add accessor methods for service globals
Phase 3 of global singleton consolidation. Add ECStore accessor
methods that delegate to process-global service singletons,
providing a unified API through ECStore.
New methods:
- notification_system() — delegates to notification_sys global
- bucket_metadata_sys() — delegates to bucket metadata global
- endpoints() — delegates to endpoints global
- region() — delegates to region global
- tier_config_mgr() — delegates to tier config global
- server_config() — delegates to server config global
- storage_class() — delegates to storage class global
1161 tests pass, clippy clean, formatting clean.
* fix: use imported functions instead of fully-qualified paths
Address review comments: use imported functions instead of
crate:: paths to avoid unused import warnings.
1160 tests pass, clippy clean.
* fix(ecstore): remove unused service imports
* fix(ecstore): skip hidden metadata in walk limit
* fix(ecstore): match walk limit to visible versions
* fix(ecstore): avoid decoding all versions for limit
* refactor(ecstore): migrate config globals into ECStore struct fields
Phase 2 of global singleton consolidation. Add server_config and
storage_class fields to ECStore, sharing the same underlying data
as the process-global LazyLock statics.
New ECStore fields:
- server_config: RwLock<Option<Config>>
- storage_class: RwLock<storageclass::Config>
New async accessor methods:
- get_server_config() / set_server_config()
- get_storage_class() / set_storage_class()
Fields initialized with defaults in ECStore::new(), synced from
globals after init() completes. Global functions preserved for
backward compatibility.
1160 tests pass, clippy clean, full workspace compiles.
* fix: address PR #3219 review comments - delegate to globals
Accessors now delegate to process-global statics to avoid state drift:
- get_server_config() delegates to get_global_server_config()
- set_server_config() updates both global and local field
- get_storage_class() delegates to get_global_storage_class()
- set_storage_class() updates both global and local field
- Removed stale sync code from init()
- Changed fields to std::sync::RwLock for sync access
1160 tests pass.
* fix: correct import order for cargo fmt
* fix: simplify config accessors - delegate to globals without local state
Remove local server_config and storage_class fields from ECStore.
Accessors now purely delegate to process-global statics, eliminating
the state drift risk identified in review.
1160 tests pass, formatting clean.
* chore(ecstore): add missing-shard first-scene diagnostics
Log rename_data quorum context behind RUSTFS_ISSUE3031_DIAG_ENABLE so partial-disk success can be correlated with later missing shard reads.
Also log put_object commit success and tmp cleanup boundaries to capture when successful quorum writes are followed by tmp_dir cleanup.
* chore(ecstore): add delete-objects lock batch diagnostics
Log DeleteObjects batch-lock request size, distributed lock quorum summary, and failed object keys behind RUSTFS_ISSUE3031_DIAG_ENABLE.
Include a Chinese operator checklist under docs/ that explains which logs to capture and how to interpret DeleteObjects lock_batch timeout incidents.
* chore: remove lock batch checklist from repo
Keep docs/issue-658-deleteobjects-lock-batch-checklist-zh.md only in the local checkout and drop it from the PR history.
* perf(erasure): remove UUID from clone + increase encode inflight budget
Two targeted optimizations for the erasure encoding hot path:
1. Erasure::clone() no longer generates Uuid::new_v4() per clone.
The _id field is unused in the hot path; reusing the original ID
eliminates a CSPRNG call per block encode (100 calls for a 100MB
object with 1MB blocks).
2. Default RUSTFS_ERASURE_ENCODE_MAX_INFLIGHT_BYTES raised from 8MB
to 32MB. This increases the encode pipeline depth from ~5 to ~20
blocks, allowing more read-ahead between the encoder and disk
writer stages. The per-request memory bound is still controlled
by the 8-block hard cap and the env var override.
3. Added encode_data_owned() utility method for zero-copy encoding
when the caller already owns a heap buffer (Vec<u8> → BytesMut
via Bytes::try_into_mut). Not used in the hot path yet but
available for future callers.
All 1157 ecstore tests pass. Criterion micro-benchmarks show no
regression (< 2% variance). Single-machine warp E2E tests were
inconclusive due to high variance; a dedicated multi-disk test
environment is needed for reliable E2E comparison.
Ref: https://github.com/rustfs/backlog/issues/659
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* chore: update Cargo.lock
* fix(erasure): align encode inflight cap and tests
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
chore(ecstore): add missing-shard first-scene diagnostics
Log rename_data quorum context behind RUSTFS_ISSUE3031_DIAG_ENABLE so partial-disk success can be correlated with later missing shard reads.
Also log put_object commit success and tmp cleanup boundaries to capture when successful quorum writes are followed by tmp_dir cleanup.
* fix(http): reduce header timeout log noise
Log peer_addr for HTTP connection errors so idle or probe traffic can be traced back to the source.
Downgrade HeaderTimeout connection logs from warn to info because these timeouts are often caused by incomplete probe traffic rather than a service-side fault.
* fix(ecstore): suppress spurious warn logs for missing trash cleanup sources
Extract `reliable_rename_inner` with a `warn_on_failure` flag so that
`rename_all_ignore_missing_source` can skip the warn log when the source
is absent. This completes the idempotent cleanup fix — the previous
implementation still emitted a `reliable_rename failed` warning before
the NotFound was caught and silenced.
Also restore the `recursive` branching in `move_to_trash` so that
non-recursive deletions use `fs::rename` directly (preserving original
semantics), while recursive deletions use the idempotent helper.
* fix(ecstore): remove dead _buf field from Erasure struct
_buf was allocated in new(), clone(), and Default but never read or written.
Each clone() was allocating block_size bytes (typically 10MB) for nothing.
- Remove _buf from struct, Default, Clone, new_with_options
- 51 erasure tests pass
* fix: remove empty line after doc comment
---------
Co-authored-by: houseme <housemecn@gmail.com>
* fix(iam): prevent transient IAM walk timeout from crashing startup
IAM startup performs a blocking full metadata walk on `.rustfs.sys/config/iam/`.
When that distributed walk times out (e.g. disk pressure after cluster reboot),
the old code treated the failure as fatal and exited the process, causing a
systemd restart loop.
Changes:
- Add `startup_iam.rs`: attempt IAM init, enter degraded mode on failure,
spawn background retry task with exponential backoff (5s→10s→20s→30s cap)
- Log level escalates to ERROR after 12 retries (~5 min) to aid diagnosis
- `/health/ready` returns 503 until IAM recovers; IAM-dependent ops return
`IamSysNotInitialized` (existing fail-closed behavior preserved)
- Fix admin path boundary matching: `/minio/administrator` no longer falsely
matches as admin prefix
- Normalize Content-Length: 0 for admin GET requests with empty body
Fixes#3175
* fix(iam): move constant assertion into const block
Fixes clippy::assertions-on-constants warning on
IAM_RETRY_ESCALATION_THRESHOLD assertion.
* fix(iam): address PR review comments
- Replace OnceLock with AtomicU64 sentinel for test isolation;
add reset_test_failure_counter() for integration tests
- Use u32::try_from() instead of `as u32` narrowing cast in
compute_backoff_interval
- Rename misleading test; update to verify finalize retry behavior
- Restructure spawn_iam_recovery_task into init-retry and
finalize-retry phases so transient readiness failures are retried
instead of leaving the server permanently degraded
* fix(iam): gate test hooks behind debug_assertions
- reset_test_failure_counter() now stores sentinel (u64::MAX) to
correctly trigger env var re-read on next call
- RUSTFS_TEST_IAM_FAIL_INIT_ATTEMPTS only honored in debug builds
- RUSTFS_TEST_IAM_RETRY_INTERVAL_MS only honored in debug builds
* test(iam): cover deferred bootstrap recovery
Add a dedicated embedded deferred-IAM integration test in a separate test binary to avoid process-global startup collisions.
Strengthen startup IAM recovery coverage with focused unit tests and keep the existing embedded smoke test isolated while carrying the manual test license header update in the same change set.
* fix(startup): tighten deferred IAM recovery path
Adopt follow-up review feedback by silencing misleading app-context warnings after IAM recovery, reusing boundary-aware path prefix checks in the readiness gate, and tying deferred IAM recovery retries to server shutdown tokens.
Keep the deferred IAM embedded integration coverage and startup recovery unit coverage green after the follow-up hardening.
* refactor(startup): simplify IAM recovery task
Collapse the deferred IAM recovery implementation back to a concrete production flow instead of keeping boxed callback seams in the runtime path.
Keep only stable backoff unit coverage in startup_iam and rely on the embedded deferred bootstrap integration test for end-to-end recovery behavior.
* refactor(startup): trim IAM recovery test scaffolding
Keep the concrete deferred IAM recovery path intact while removing bulky test-only async loop scaffolding from startup_iam.
Retain the stable backoff unit checks and rely on the embedded deferred bootstrap integration test for end-to-end recovery coverage.
* fix: apply code review improvements from PR #3188 review
- Simplify RecoveryFuture type alias by removing unnecessary lifetime
- Fix finalize_iam_recovery to return Err if app context unavailable
- Update bootstrap_or_defer_iam_init doc comment to reflect Err case
- Use boundary-aware has_path_prefix for admin path matching in utils.rs
- Add test for adminx boundary rejection in utils.rs and layer.rs
- Improve embedded deferred IAM test with timeout wrapper
* style: merge has_path_prefix import into existing use block
* fix(iam): address final review follow-ups
- fix main startup readiness publication to pass ServiceStateManager correctly
- centralize IAM test env keys in rustfs_config and reuse them in runtime/tests
- keep deferred IAM bootstrap validation aligned with the final review fixes
* fix: isolate listing timeouts from drive health
Keep walk_dir scanner timeouts request-scoped instead of marking local drives faulty.
Add regression coverage for follow-up bucket info, set-level list_path, and system-prefix listings after prior walk timeouts.
* test(iam): gate deferred bootstrap test to debug
Align the deferred IAM embedded integration test with debug-only IAM fault injection hooks so release-profile runs do not assert deferred bootstrap behavior that cannot be triggered.
* test(ecstore): bound prior walk timeout regressions
- set walk_dir stall timeout explicitly in prior-timeout listing tests
- keep the system-prefix follow-up listing scoped to the same base dir
- assert the expected directory entry so the timeout regression test stays fast and stable
* fmt
* fix(ecstore): retry transient walk dir stream errors
Retry RemoteDisk walk_dir once when the first failure looks transport-related so restart-time HttpReader stream errors do not immediately count as hard listing failures.
Improve HttpReader request and stream error messages with method and URL context, and add regression coverage for retry recovery and diagnostics.
* fix(ecstore): avoid retry after partial walk dir stream
Stop retrying walk_dir after copy_stream_with_buffer has already written partial bytes into the destination writer.
Keep the retry only for open_walk_dir failures and add a regression test that proves partial stream failures are returned without issuing a second stream.
* feat(ecstore): add object lock diagnostics
Add configurable namespace lock diagnostics for object operations so production contention can be traced by operation, owner, and key.
Wrap object read/write lock acquisition in diagnostic guards across get, head, put, delete, copy, and multipart flows, and log slow acquisition and long hold durations behind new RUSTFS_OBJECT_LOCK_DIAG_* settings.
Verification:
- make pre-commit
* feat(obs): expose object lock diagnostics metrics
Add Prometheus metrics and Grafana panels for object namespace lock diagnostics, covering slow acquire counts, slow hold counts, acquire duration, hold duration, and the diagnostics-enabled state.
Adopt PR review feedback by keeping diagnostic guards alive through the guarded operation so long-hold warnings and metrics are emitted, reusing shared env helpers, and reducing default-path overhead when diagnostics are disabled.
Verification:
- cargo check -p rustfs-ecstore -p rustfs-io-metrics
- cargo test -p rustfs-ecstore store::object -- --nocapture
- make pre-commit
* perf(obs): reduce object lock diag overhead
Avoid repeated environment parsing on hot object-lock paths by caching the diagnostics-enabled flag, and stop allocating label strings for object lock metrics by recording static labels directly.
Strengthen io-metrics tests by using a local recorder and asserting that the expected object lock diagnostic counters, gauges, and histograms are emitted.
Verification:
- cargo check -p rustfs-ecstore -p rustfs-io-metrics
- cargo test -p rustfs-io-metrics -- --nocapture
- make pre-commit
* fix(rio): surface incomplete put bodies
Propagate incomplete PUT request bodies as IncompleteBody instead of allowing erasure encode to treat truncated input as a normal EOF.\n\n- mark premature EOFs in HardLimitReader with an explicit IncompleteBody error\n- preserve EOF error chains through read_full and map them to S3 IncompleteBody\n- stop erasure encode from swallowing UnexpectedEof on truncated input\n- add regression tests for reader, erasure encode, and API error mapping\n\nRefs: rustfs/backlog#654
* fix(s3): honor decoded length for aws chunked put
Use x-amz-decoded-content-length for aws-chunked PutObject requests so trailer-checksum uploads are sized against the decoded payload instead of the wire-encoded content-length.\n\n- prefer decoded content length for aws-chunked put bodies\n- add a regression test covering the size selection logic\n- keeps the incomplete body fix working for truly truncated uploads while restoring checksum trailer compatibility\n\nRefs: rustfs/backlog#654
* fix(io): follow up review comments on incompletebody handling
Address PR review feedback by restoring read_full's existing EOF contract, adding a dedicated read_full_or_eof helper for erasure encoding, covering nested incomplete-body error chains, and documenting plus hardening aws-chunked size selection.\n\n- keep read_full returning early EOF on empty reads\n- use read_full_or_eof only in erasure encoding paths\n- detect aws-chunked via content-encoding or transfer-encoding\n- add nested error-chain and aws-chunked regression tests\n\nRefs: rustfs/backlog#654
* fix(rio): surface incomplete put bodies
Propagate incomplete PUT request bodies as IncompleteBody instead of allowing erasure encode to treat truncated input as a normal EOF.\n\n- mark premature EOFs in HardLimitReader with an explicit IncompleteBody error\n- preserve EOF error chains through read_full and map them to S3 IncompleteBody\n- stop erasure encode from swallowing UnexpectedEof on truncated input\n- add regression tests for reader, erasure encode, and API error mapping\n\nRefs: rustfs/backlog#654
* fix(s3): honor decoded length for aws chunked put
Use x-amz-decoded-content-length for aws-chunked PutObject requests so trailer-checksum uploads are sized against the decoded payload instead of the wire-encoded content-length.\n\n- prefer decoded content length for aws-chunked put bodies\n- add a regression test covering the size selection logic\n- keeps the incomplete body fix working for truly truncated uploads while restoring checksum trailer compatibility\n\nRefs: rustfs/backlog#654
* fix(io): follow up review comments on incompletebody handling
Address PR review feedback by restoring read_full's existing EOF contract, adding a dedicated read_full_or_eof helper for erasure encoding, covering nested incomplete-body error chains, and documenting plus hardening aws-chunked size selection.\n\n- keep read_full returning early EOF on empty reads\n- use read_full_or_eof only in erasure encoding paths\n- detect aws-chunked via content-encoding or transfer-encoding\n- add nested error-chain and aws-chunked regression tests\n\nRefs: rustfs/backlog#654
* fix(rio): reject bytes beyond hard limit
* fix(ecstore): reject zero-sized erasure blocks
* fix(rebalance): require target goal before completion
Stop treating pools within a fixed tolerance of the rebalance goal as completed before any data has moved.
Add a regression test for the two-pool imbalance reported in issue 3137.
* chore(rebalance): clarify goal completion wording
---------
Co-authored-by: Henry Guo <marshawcoco@users.noreply.github.com>
Co-authored-by: loverustfs <hello@rustfs.com>
Co-authored-by: houseme <housemecn@gmail.com>
* feat(admin): restore config admin compatibility
Co-authored-by: weisd <im@weisd.in>
* fix(admin): align config admin clean rebuild
Co-authored-by: weisd <im@weisd.in>
* fix(admin): align config history and peer signals
* fix(admin): harden config admin mutations
* fix(admin): tighten config review follow-ups
* perf(admin): reuse env snapshot in config render
* fix(ecstore): clean up config admin and listing error handling
Remove redundant is_all_volume_not_found check in list_merged, add
storage class encode/decode roundtrip tests, fresh boot integration
test, and config admin clean rebuild improvements.
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(admin): sync global server config on mutation and reload
Change GLOBAL_SERVER_CONFIG from OnceLock to RwLock so config mutations
(set/del/restore/reload) are visible to readers without restart. Call
set_global_server_config after every store save and on snapshot reload.
Register storage_class as a dynamic config subsystem.
Co-authored-by: hehutu <heihutu@gmail.com>
* style: apply rustfmt to config and admin tests
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(test): update signal_service test for storage_class dynamic subsystem
storage_class is now a valid dynamic config subsystem, so the
"requires object layer" test should expect "storage layer not initialized"
instead of "unsupported dynamic config subsystem".
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(config): publish storage_class runtime config on dynamic reload
Change GLOBAL_STORAGE_CLASS from OnceLock to RwLock so runtime updates
are possible. apply_storage_class_runtime_config now actually publishes
the parsed config via set_global_storage_class instead of dropping it.
Addresses review feedback: storage_class was marked as dynamically
applied but the parsed result was discarded, so mc admin config set
returned config_applied=true while the runtime kept using stale parity
settings until restart.
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(admin): harden config init, history ordering, and env redaction
- Change GLOBAL_SERVER_CONFIG from RwLock<Config> to RwLock<Option<Config>>
initialized with None, preserving "not initialized" detection via None
- Move save_server_config_history before save_server_config_to_store in
SetConfigKVHandler, DelConfigKVHandler, and SetConfigHandler so a
restore point exists before mutations are persisted
- Redact sensitive env override values with *redacted* instead of
silently omitting the line, improving admin visibility
- Add code comment explaining VolumeNotFound removal rationale in
list_merged for listing paths
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(config): keep in-memory config in sync after set/restore/reload
GLOBAL_SERVER_CONFIG was a OnceLock set once at startup and never
updated. After mc admin config set writes to the store, any fallback
to get_global_server_config() returned stale init-time data. Similarly,
reload_runtime_config_snapshot read from the store but discarded the
result.
- Replace OnceLock with RwLock for GLOBAL_SERVER_CONFIG and
GLOBAL_STORAGE_CLASS so they can be updated at runtime
- Add set_global_server_config / set_global_storage_class setters
- Call set_global_server_config after every config save (set-kv,
del-kv, set-config, restore-history)
- Re-apply dynamic subsystems (storage_class, audit_webhook,
audit_mqtt) and signal peers in reload_runtime_config_snapshot
and full-config operations
- Fix render_selected_config scope boundary check: track per-scope
line count instead of checking global lines.is_empty()
- Include STORAGE_CLASS_SUB_SYS in is_dynamic_config_subsystem so
apply_storage_class_runtime_config is reachable
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(storageclass): use CLASS_RRS key in lookup_config for RRS parity
lookup_config used kvs.get(RRS) where RRS="REDUCED_REDUNDANCY", but the
admin config path writes the key as CLASS_RRS="rrs". This caused RRS
values to never be read back, always falling back to default parity.
- Changed kvs.get(RRS) to kvs.get(CLASS_RRS) in lookup_config
- Added regression tests verifying RRS read/write consistency
Co-authored-by: hehutu <heihutu@gmail.com>
* fix(config): add peer-side logging and don't swallow apply errors
- Add tracing::warn! in reload_dynamic_config_runtime_state and
reload_runtime_config_snapshot when config read or subsystem apply
fails, so on-host diagnostics show which signal failed and why
- Change `let _ = apply_dynamic_config_for_subsystem(...)` to
`if let Err(err) = ... { warn!(...) }` in reload_runtime_config_snapshot
so per-subsystem failures are logged instead of silently swallowed
- Remove weak test global_server_config_returns_none_before_init that
had no meaningful assertion due to shared global state
Co-authored-by: hehutu <heihutu@gmail.com>
* style: apply rustfmt to config and storageclass tests
Co-authored-by: hehutu <heihutu@gmail.com>
---------
Co-authored-by: weisd <im@weisd.in>
Co-authored-by: hehutu <heihutu@gmail.com>
* perf: reduce spawn_blocking contention in PUT path (~23% throughput gain)
Flame graph profiling identified tokio blocking pool mutex contention
as the #1 bottleneck (17.3% of CPU time). Each spawn_blocking call must
acquire parking_lot::raw_mutex to enqueue work. With 16 concurrent PUTs
× 4 disks × 3+ spawn_blocking per disk, this became a serialization point.
Optimizations applied:
- Merge make_dir_all + file write into single spawn_blocking
- Merge read_file + parse + write + rename into single spawn_blocking
for inline objects (small files)
- Optimize reliable_rename to try rename first, mkdir only on ENOENT
- Optimize remove/remove_std to try remove_file first, EISDIR fallback
- Add encode_inline_small fast path for small objects
- Parallelize bitrot writer creation with join_all
Benchmark (4KiB PUT, 4-disk EC, 16 concurrent, 8 rounds, randomized A/B):
Baseline: ~950 obj/s → Optimized: ~1173 obj/s (+23%)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: address review comments
- Return old_data_dir for non-inline rename_data path (was incorrectly None)
- Restore delete_all cleanup of PUT temp data on failure paths
- Fix cargo fmt formatting
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* style: apply rustfmt from stable 1.96.0
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: collapse nested if-let chains for clippy compliance
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: address Copilot review comments
- fs.rs: handle macOS EPERM from remove_file on directories
- os.rs: restore NotFound=Ok(()) semantics on first rename attempt
- local.rs: use try-rename-then-mkdir pattern for inline rename_data
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* test: add unit tests for encode_inline_small fast path
* test: fix comment and line length in encode_inline_small tests
* fix: revert reliable_rename and write_all_internal to match original
Restore the original reliable_rename logic (check parent exists, then
rename in loop) and the original write_all_internal (make_dir_all outside
spawn_blocking). The optimization changes caused a CI-only test failure
in capacity_dirty_scope_test that could not be reproduced locally.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: update comment and fix formatting for CI
- Fix macOS/BSD comment to accurately say macOS only
- Fix encode_inline_small test formatting to match rustfmt 1.96.0
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix: propagate inline rename errors
* fix: retry rename when parent missing
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
* fix: preserve S3 listing pagination parity
S3 listing responses need stable wire semantics across encoded listings, version markers, and quorum-sensitive metadata reads. This tightens marker handling, response encoding, and version-marker pagination while keeping the changes scoped to listing paths and regression tests.
Constraint: Branch, code, commit, and PR text must avoid restricted upstream project naming.
Constraint: Verification required Rust 1.95 toolchain path because Homebrew cargo-clippy resolved to 1.94.
Rejected: Treat null version-id-marker as no marker | repeats or skips the marker boundary for null-version listings.
Rejected: Compare encoded next-marker values directly | encoded output can diverge from raw listing order.
Confidence: high
Scope-risk: moderate
Directive: Do not change listing marker semantics without covering V1, V2, versions, encoded responses, and null-version markers together.
Tested: cargo fmt --all --check
Tested: cargo clippy --workspace --all-features --all-targets -- -D warnings
Tested: make pre-commit with LC_ALL=en_US.UTF-8 and Rust 1.95 PATH
Tested: git diff --check
Not-tested: live distributed object-store compatibility test against a remote cluster
* fix: keep listing echo fields raw
S3 compatibility tests expect list response echo fields such as Prefix, Delimiter, StartAfter, and Marker to preserve the request value even when URL encoding is requested. Keep URL encoding scoped to object keys and common prefixes while preserving the next-marker raw comparison fix.
Constraint: PR branch was updated with latest main before this fix.
Constraint: Do not use restricted upstream project naming in commit text.
Rejected: Encode all response string fields | breaks compatibility tests for unreadable prefix values.
Confidence: high
Scope-risk: narrow
Tested: cargo test -p rustfs list_objects --lib
Tested: cargo fmt --all --check
Tested: cargo clippy --workspace --all-features --all-targets -- -D warnings
Tested: make pre-commit
Tested: git diff --check
Not-tested: live CI s3 compatibility rerun before push
* fix: accept empty listing continuation token
S3 compatibility tests treat an empty ListObjectsV2 continuation token as an explicit empty echo value, not as an invalid base64 token. Preserve the request echo while skipping decoded-token pagination for the empty string.
Constraint: Keep invalid non-empty continuation tokens rejected before store lookup.
Rejected: Drop empty continuation tokens from the response | compatibility tests assert the empty echo field is present.
Confidence: high
Scope-risk: narrow
Tested: cargo test -p rustfs list_objects --lib
Tested: cargo fmt --all --check
Tested: cargo clippy --workspace --all-features --all-targets -- -D warnings
Tested: make pre-commit
Tested: git diff --check
Not-tested: live CI s3 compatibility rerun after this push
* fix: delete explicit null object versions
S3 compatibility cleanup lists unversioned objects with VersionId=null and then sends that value back through DeleteObjects. The API layer keeps null as an internal sentinel, but the storage delete path must map it back to the stored null version instead of treating it as a real UUID.
Constraint: Keep the wire response able to echo null version IDs.
Rejected: Treat VersionId=null the same as an absent version id everywhere | versioned and suspended buckets need explicit null version semantics.
Confidence: high
Scope-risk: moderate
Tested: cargo test -p rustfs-ecstore delete_file_info_version_id_maps_explicit_null_version_to_stored_null
Tested: cargo test -p rustfs normalize_delete_objects_version_id_preserves_explicit_null_marker --lib
Tested: cargo fmt --all --check
Tested: cargo clippy --workspace --all-features --all-targets -- -D warnings
Tested: make pre-commit
Tested: git diff --check
Not-tested: live CI s3 compatibility rerun after this push
* fix: keep version marker scoped to marker key
Version listing cleanup can delete the key returned as the previous page marker before requesting the next page. When the marker key is no longer present, the version marker must not be applied to the first later key, otherwise each cleanup page can skip one null-version object and leave the bucket non-empty.
Constraint: Preserve version-marker pagination for marker keys that still exist.
Rejected: Drop version markers whenever the marker key was supplied | multi-version marker keys still need intra-key pagination.
Confidence: high
Scope-risk: narrow
Tested: cargo test -p rustfs-ecstore version_marker_is_applied_only_when_key_marker_entry_is_present
Tested: cargo test -p rustfs-ecstore delete_file_info_version_id_maps_explicit_null_version_to_stored_null
Tested: cargo test -p rustfs normalize_delete_objects_version_id_preserves_explicit_null_marker --lib
Tested: cargo fmt --all --check
Tested: cargo clippy --workspace --all-features --all-targets -- -D warnings
Tested: make pre-commit
Tested: git diff --check
Not-tested: live CI s3 compatibility rerun after this push
---------
Co-authored-by: houseme <housemecn@gmail.com>
Co-authored-by: loverustfs <hello@rustfs.com>
* fix(ecstore): send valid ping body in remote locker
Build ping requests with a flatbuffer payload so health checks remain compatible with the ping response parser after restart.
* fix(bench): use multi-host warp target during failover
Normalize comma-separated warp host lists in run_object_batch_bench and let four-node failover bench pass BENCH_WARP_HOSTS so rolling restart does not pin load to a single restarting node.
* feat(health): add compat health probes with busy/KMS checks
- Add /health/live liveness probe endpoint
- Add busy protection (429) for readiness probes, gated by RUSTFS_HEALTH_COMPAT_BUSY_CHECK_ENABLE
- Add KMS readiness check for /health/ready, gated by RUSTFS_HEALTH_COMPAT_KMS_READY_CHECK_ENABLE
- Add lock quorum status caching with TTL to reduce RPC pressure
- Consolidate health response building into build_health_response_parts
- Register /health/live in console router and readiness gate
- Remove MinIO references from newly added health code
* fix(health): decouple kms readiness from lock quorum
Treat empty ping bodies as liveness probes, add a startup cleanup barrier for early walk_dir calls, and delay immediate background cleanup/capacity timers to reduce transient restart-time VolumeNotFound noise.
Also downgrade expected missing-path producer results during startup from generic errors to warnings while preserving existing storage semantics.
* fix(ecstore): harden issue3031 multipart validation path
- clear stale multipart part destinations before rename fan-out
- add repeated part overwrite regression coverage
- reduce remote disk startup false-fault escalation to suspect-first
- refine remote locker diagnostics and lower scanner leader-lock log noise
- add a dedicated 4-node issue3031 docker validation script
* refactor(admin): inline console version json macro
- drop the unused serde_json::json import in admin console
- call serde_json::json! inline in version_handler
- keep the console version response behavior unchanged
* fix(remote-disk): recover suspect health on probe success
- record probe success during remote disk health checks so suspect drives recover
- use async_with_vars for the remote disk health probe test
- make the missing-listener test assert the state transition more robustly
* fix(readiness): gate on lock quorum health
Include distributed lock quorum in runtime readiness decisions and expose the signal in health responses.
Map namespace lock quorum failures to ServiceUnavailable instead of InternalError so clients can safely retry while keeping quorum enforcement unchanged.
* fix(ecstore): restore namespace lock error decoding
Add the missing from_u32 arm for NamespaceLockQuorumUnavailable and cover the numeric roundtrip in error tests.
Henry Guo