 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
|
0e234f5c80
|
feat: Phase 10 CP10-1 -- NVMe/TCP target MVP, 109 tests
NVMe over Fabrics (TCP) target implementation sharing the same BlockVol
engine, fencing, replication, and failover as the existing iSCSI target.
New package: weed/storage/blockvol/nvme/ (11 files, 2,242 production LOC)
- protocol.go: PDU types, opcodes, status codes, marshal/unmarshal
- wire.go: TCP reader/writer with header bounds validation
- controller.go: IC handshake, per-queue state, command dispatch, KATO
- fabric.go: Connect (admin+IO), PropertyGet/Set, Disconnect
- identify.go: Controller/Namespace/NS list/NS descriptors (Linux 5.15)
- admin.go: SetFeatures, GetFeatures, GetLogPage (SMART/ANA), KeepAlive
- io.go: Read (C2HData), Write (inline), Flush, WriteZeros/Trim
- server.go: TCP listener, admin session registry, graceful shutdown
- adapter.go: BlockVol-to-NVMe bridge, error mapping, ANA state
Integration: NVMeConfig + CLI flags (-block.nvme.*), disabled by default.
Key design: inline-data writes only (no R2T), MaxH2CDataLength=32KB,
single ANA group coherent with BlockVol role, CNTLID session registry
for cross-connection IO queues, HostNQN continuity enforcement.
Tests: 65 dev + 44 QA adversarial = 109 total, all passing.
Bugs fixed during review: IO queue cross-connection (A), header bounds
validation (B), write payload size check (C), disconnect error (D),
stream desync prevention (E), HostNQN enforcement (F), capsule-before-IC
state guard (H), flowCtlOff SQHD timing (I).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-03-08 16:52:37 -07:00 |
|