mirrors/seaweedfs
1
0
Fork 0
mirror of https://github.com/seaweedfs/seaweedfs.git synced 2026-09-19 22:44:18 +00:00
Code Issues Packages Projects Releases Wiki Activity
13,152 Commits 181 Branches 360 Tags
3da4c19046e06ec7e75cb449e0d9287ef1a3473b
Commit Graph
8 Commits
Author SHA1 Message Date
pingqiuandClaude Opus 4.6 ebe95b6e2e fix: flusher OOM on multi-block writes + testrunner enhancements 
Bug: flusher.go:336 allocated make([]byte, entryLen) per dirty block
instead of per unique WAL entry. A 4MB WriteLBA creates 1024 dirty map
entries (one per 4KB block), all sharing the same WAL offset. The flusher
read the full 4MB WAL entry 1024 times into separate buffers:
1024 × 4MB = 4GB per 4MB write → OOM on mkfs.ext4.

Root cause: flusher assumed 1:1 dirty-block-to-WAL-entry mapping.
WriteLBA supports multi-block writes but the flusher never deduplicated
shared WAL offsets.

Fix: deduplicate WAL reads by WalOffset in flushOnceLocked(). Multiple
dirty blocks from the same WAL entry share one read buffer and one
DecodeWALEntry call. Memory: O(WAL_entries × size) not O(blocks × size).
For a 4MB write: 4GB → 4MB.

Verified on hardware (m01/M02 25Gbps RoCE):
- Before: mkfs.ext4 → VS RSS 100MB→25GB → OOM killed
- After: mkfs.ext4 → VS RSS 129MB stable, mkfs succeeds
- pgbench TPC-B c=4: 1,248 TPS (RF=1, previously blocked by OOM)

Tests added:
- flusher_test.go: flush_multiblock_shared_wal_read (16 blocks share
  one WAL offset, flush dedup verified)
- flusher_test.go: flush_multiblock_data_correct (3 mixed multi-block
  writes, all data correct after flush)
- test/component/large_write_test.go: 7 component tests (single 4MB,
  sequential mkfs sim, concurrent, mixed sizes, production volume,
  flusher throughput 30s sustained)
- iscsi/large_write_mem_test.go: 2 iSCSI session memory tests (4MB
  R2T flow, slow device)

Testrunner enhancements (same commit — all tested on hardware):
- discover_primary action: maps primary IP → topology node name,
  supports alt_ips for multi-NIC (RoCE + management)
- NodeSpec.AltIPs field for multi-NIC node identification
- 5 new YAML scenarios: ec3, ec5, degraded sync_all/best_effort, pgbench
- All 13 hardware-verified scenarios PASS

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-02 14:24:10 -07:00
pingqiuandClaude Opus 4.6 785a7d7efd feat: wire real pinner into flusher retention + real WAL scan executor (Phase 07 P1) 
Pinner wired to real retention:
- NewPinner calls vol.SetV2RetentionFloor(p.MinWALRetentionFloor)
- Flusher.RetentionFloorFn() / SetRetentionFloorFn() exposed
- SetV2RetentionFloor chains with existing shipper retention floor
- Holds actually prevent WAL reclaim (not just tracked state)

Executor uses real WAL scan:
- BlockVol.ScanWALEntries(fromLSN, callback) wraps wal.ScanFrom
  with real fd, walOffset, checkpointLSN
- Executor.StreamWALEntries uses ScanWALEntries (not stub)
- Reads real WAL entries, tracks highest LSN scanned

CommittedLSN mapping:
- Explicitly documented as interim V1 model (committed = checkpointed)
- Will diverge when V2 distributed commit separates from local flush

Carry-forward:
- TransferSnapshot/TransferFullBase/TruncateWAL: stubs (need extent I/O)
- Control intent from confirmed failover: deferred

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-30 20:01:46 -07:00
Ping QiuandClaude Opus 4.6 075ff52219 feat: CP11B-3 safe ops — promotion hardening, preflight, manual promote 
Six-task checkpoint hardening the promotion and failover paths:

T1: 4-gate candidate evaluation (heartbeat freshness, WAL lag, role,
    server liveness) with structured rejection reasons.
T2: Orphaned-primary re-evaluation on replica reconnect (B-06/B-08).
T3: Deferred timer safety — epoch validation prevents stale timers
    from firing on recreated/changed volumes (B-07).
T4: Rebuild addr cleanup on promotion (B-11), NVMe publication
    refresh on heartbeat, and preflight endpoint wiring.
T5: Manual promote API — POST /block/volume/{name}/promote with
    force flag, target server selection, and structured rejection
    response. Shared applyPromotionLocked/finalizePromotion helpers
    eliminate duplication between auto and manual paths.
T6: Read-only preflight endpoint (GET /block/volume/{name}/preflight)
    and blockapi client wrappers (Preflight, Promote).

BUG-T5-1: PromotionsTotal counter moved to finalizePromotion (shared
    by both auto and manual paths) to prevent metrics divergence.

24 files changed, ~6500 lines added. 42 new QA adversarial tests.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-13 17:21:17 -07:00
Ping QiuandClaude Opus 4.6 1b3edd7856 feat: CP11A-2 coordinated expand protocol for replicated block volumes 
Two-phase prepare/commit/cancel protocol ensures all replicas expand
atomically. Standalone volumes use direct-commit (unchanged behavior).

Engine: PrepareExpand/CommitExpand/CancelExpand with on-disk
PreparedSize+ExpandEpoch in superblock, crash recovery clears stale
prepare state on open, v.mu serializes concurrent expand operations.

Proto: 3 new RPCs (PrepareExpand/CommitExpand/CancelExpandBlockVolume).

Coordinator: expandClean flag pattern — ReleaseExpandInflight only on
clean success or full cancel. Partial replica commit failure calls
MarkExpandFailed (keeps ExpandInProgress=true, suppresses heartbeat
size updates). ClearExpandFailed for manual reconciliation.

Registry: AcquireExpandInflight records PendingExpandSize+ExpandEpoch.
ExpandFailed state blocks new expands until cleared.

Tests: 15 engine + 4 VS + 10 coordinator + heartbeat suppression
regression + updated QA CP82/durability tests with prepare/commit mocks.

Also includes CP11A-1 remaining: QA storage profile tests, QA
io_backend config tests, testrunner perf-baseline scenarios and
coordinated-expand actions.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-12 15:06:48 -07:00
Ping QiuandClaude Opus 4.6 4c5f9f2b9d feat: CP10B-1 NVMe/TCP RX/TX split + CP10B-2 bench/profiling fixes 
RX/TX split: rxLoop reads PDUs, txLoop writes responses via respCh.
Handlers refactored to void + enqueueResponse pattern. IOCCSZ fix
enables inline write data (100K IOPS vs 15K before). R2T deadlock
fix via completeWaiters. Shutdown cleans up pendingCapsules buffers.

Bench: ParseFioMetric accepts plain/quoted numbers for aggregated
medians. Profiling actions: pprof_capture, vmstat_capture, iostat_capture.

196 NVMe tests, 92 testrunner actions.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-10 15:09:41 -07:00
Ping QiuandClaude Opus 4.6 3557ae283f feat: Phase 10 CP10-3 -- NVMe/TCP Tier 1 optimizations, WAL admission control, benchmark platform 
CP10-3 Tier 1 optimizations (T1-T4):
- TCP_NODELAY + 256KB socket buffers on NVMe/TCP connections
- Response batching: all C2H data chunks + CapsuleResp in single flush
- Tiered buffer pool (4KB/64KB/256KB sync.Pool) for write payloads
- Configurable MaxH2CDataLength wiring through controller/IC/chunking

BUG-CP103-1: NVMe write retry with jittered backoff for transient WAL pressure
- writeWithRetry() with bounded backoff [50/200/800ms]
- throttleOnWALPressure() pre-write delay above 90% WAL usage
- WALPressureProvider interface + NVMeAdapter.WALPressure()

BUG-CP103-2: Volume-level WAL admission control
- WALAdmission with counting semaphore (max concurrent writers)
- Soft watermark (0.7): small delay to desynchronize herd
- Hard watermark (0.9): block until flusher drains
- Single-deadline budget shared across watermark wait + semaphore
- Close-aware during both watermark and semaphore waits
- Wired into BlockVol.WriteLBA() and Trim()

Benchmark platform enhancements:
- NVMe benchmark actions and scenarios (A/B, CW sweep, IOQ sweep)
- Database benchmark actions (SQLite, pgbench)
- K8s operator QA reconciler tests
- New testrunner scenarios for HA, fault injection, CSI lifecycle

Test counts: 213 NVMe + 625 engine + operator + testrunner tests, all passing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-09 17:44:01 -07:00
Ping QiuandClaude Opus 4.6 9acd187587 feat: Phase 8 complete -- CP8-5 stability gate, lease grant fix, Docker e2e, 13 chaos scenarios 
Phase 8 closes with all 6 checkpoints done (CP8-1 through CP8-5 + CP8-3-1):
- CP8-5: 12/12 enterprise QA scenarios PASS on real hardware (m01/M02)
- Master-authoritative lease grants (BUG-CP85-11): master renews primary
  write leases on every heartbeat response, replacing retain-until-confirmed
  assignment queue semantics that caused 30s lease expiry
- Post-rebuild WAL shipping gap fix (BUG-CP85-1): syncLSNAfterRebuild
  advances replica nextLSN so WAL entries are accepted after rebuild
- Block heartbeat startup race fix (BUG-CP85-10): dynamic blockService
  check on each tick instead of one-shot at loop start
- 8 new tests: 4 engine lease grant + 4 registry lease grant
- 13 new YAML scenarios: chaos (kill-loop, partition, disk-full),
  database integrity (sqlite crash, ext4 fsck), perf baseline,
  metrics verify, snapshot stress, expand-failover, session storm,
  role flap, 24h soak
- 12 new testrunner actions (database, fsck, grep_log, write_loop_bg,
  stop_bg, assert_metric_gt/eq/lt) + phase repeat support
- Docker compose setup + getting-started guide for block storage users
- 960+ cumulative unit tests, 24 YAML scenarios

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-07 21:30:14 -08:00
Ping QiuandClaude Opus 4.6 da1b81d1c9 feat: CP8-3-1 durability modes + testrunner platform + 21 adversarial tests 
Durability mode implementation (sync_all, sync_quorum, best_effort):
- DurabilityMode type with superblock persistence, parse/validate/string
- MakeDistributedSync mode-aware barrier enforcement in dist_group_commit
- blockerr sentinel package (ErrDurabilityBarrierFailed, ErrDurabilityQuorumLost)
- gRPC create path: mode validation, idempotent create consistency, partial cleanup
- F1: strict mode rejects partial replica provisioning with cleanup
- F3: empty heartbeat does not overwrite persisted strict mode
- F4: SCSI error mapping uses errors.Is sentinels (not string matching)
- Proto/wire/blockapi/CLI/UI plumbing for durability_mode field
- Observability dashboard: cluster health cards + per-volume columns

Testrunner platform (YAML-driven integration test framework):
- Engine, parser, registry, reporter (JUnit XML + HTML), metrics scraping
- 52 registered actions: block, iSCSI, I/O, fault injection, assertions
- Baseline regression framework with 7 hard-fail conditions
- 15 YAML scenarios (smoke, crash, HA, fault, consistency, snapshot)
- 49 unit tests for testrunner internals

QA adversarial suite (21 tests, all PASS):
- Idempotent create mode/RF mismatch detection
- Heartbeat mode downgrade prevention (F3)
- sync_all/sync_quorum partial replica enforcement (F1)
- Concurrent create race safety
- Failover/expand mode preservation
- Cleanup resilience when delete fails
- Master restart auto-register mode handling
- Superblock roundtrip all 3 modes
- Validate edge cases (mode×RF matrix)
- RequiredReplicas quorum math verification
- Sentinel error categorization

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
 2026-03-06 01:06:51 -08:00