Commit Graph

5749 Commits

Author SHA1 Message Date
overtrue b91cb365c4 fix(ecstore): remove redundant clone in test 2026-08-23 09:50:28 +08:00
overtrue 32436e322a fix(ecstore): clear rio-v2 lint errors 2026-08-23 05:27:57 +08:00
overtrue 12f56c9ebd fix(ecstore): fence decommission cancellation 2026-08-23 05:22:29 +08:00
overtrue acc37b49c8 test(ecstore): cover final sweep cancel fence 2026-08-23 03:21:15 +08:00
overtrue 2379bfb5a3 fix(ecstore): fence final decommission sweep 2026-08-23 03:15:33 +08:00
overtrue 4527b003cc fix(ecstore): remove duplicate decommission error helper 2026-08-23 03:00:41 +08:00
overtrue c83aa7f974 Merge remote-tracking branch 'origin/main' into overtrue/fix-1911-ilm-metadata-decommission
# Conflicts:
#	crates/ecstore/src/core/pools.rs
2026-08-23 02:58:32 +08:00
Zhengchao An 2d7120460b test(e2e): remove fake KMS suite results (#6401) 2026-08-23 01:45:15 +08:00
Zhengchao An b91845c98c ci: align cache writer and reader keys (#6398) 2026-08-23 01:44:50 +08:00
Zhengchao An 7ba6f8cb33 test(e2e): fix cluster nightly oracles (#6397) 2026-08-23 01:44:35 +08:00
Zhengchao An f44b30c61a ci(perf): fix nightly regression baseline (#6389) 2026-08-23 01:43:52 +08:00
Zhengchao An c6590182ed ci(mint): pin manual image default (#6387) 2026-08-23 01:43:27 +08:00
Zhengchao An 6d85a9c6a8 ci(s3tests): stabilize HAProxy request handling (#6386) 2026-08-23 01:42:58 +08:00
Zhengchao An 26e6508b64 fix(ecstore): reject equal-time latest identity conflicts before index fallback (#6374) 2026-08-23 01:42:11 +08:00
GatewayJ 51e369be6c fix(policy): accept legacy bucket policy ID field (#6362) 2026-08-23 01:40:44 +08:00
Zhengchao An ddc4120c82 ci: detect incomplete and stale scheduled validations (#6357) 2026-08-23 01:40:28 +08:00
cxymds 87235ffd28 perf(ecstore): bound decommission entry workers (#6360) 2026-08-23 01:40:03 +08:00
cxymds 62c465ecef fix(heal): supervise scheduler task panics (#6351) 2026-08-23 01:39:06 +08:00
houseme f1b92af4a3 feat(ecstore): coalesce GET ReadVersion RPCs (#6395) 2026-08-23 01:15:38 +08:00
Zhengchao An 7c1a76dfd9 fix(ecstore): fence deletes against decommission commits (#6363)
* fix(ecstore): fence deletes against decommission commits

* fix(ecstore): preserve decommission target write locks

* fix(ecstore): preserve delete markers during source cleanup

* fix(ecstore): route batch delete markers to active pools

* fix(ecstore): preserve batch delete pool errors

* fix(ecstore): retain source-set lock during cleanup

* test(ecstore): exercise decommission delete fences

* test(ecstore): finish decommission delete fence scenario

* fix(ecstore): reuse fixed fence for reverse decommission

* fix(ecstore): fence decommission commit loss

* fix(ecstore): annotate batch delete fallback

* test(ecstore): fix decommission fence fixtures

* fix(ecstore): unblock decommission delete fences

* fix(ecstore): preserve distributed decommission set locks

* fix(ecstore): match decommission lock backend domain

* test(ecstore): align decommission fence barriers

* fix(ecstore): satisfy delete fence lint checks

* fix(rebalance): preserve access-denied delete errors
2026-08-23 00:40:38 +08:00
Zhengchao An 3ddf1a81ac test(kms): replace 33 hard-coded startup sleeps with readiness probe (#6349)
* refactor(e2e/kms): replace fixed startup sleeps with KMS readiness probe

Replace 33 hard-coded sleep(3s) / sleep(2s) startup waits in KMS e2e tests
with an active readiness probe (wait_for_kms_ready) that polls the KMS
status endpoint with exponential backoff (200ms→1s, 5s budget).

This cuts per-test startup latency from a fixed 3s to ~200-500ms while
remaining robust against slow CI machines.

Non-startup sleeps (ILM polling loops, fault-recovery detection delays,
test-runner inter-test pauses) are left untouched.

* style: cargo fmt

* fix(kms): use .expect() instead of ? in test functions that return ()

7 call sites of wait_for_kms_ready() used ? in async test functions
that return () instead of Result. Changed to .expect("KMS ready").

* fix(kms): enforce readiness probe deadline

* fix(kms): validate readiness backend status
2026-08-22 16:35:57 +00:00
Zhengchao An cc412914d5 feat(connect): emit durable heartbeats (#6383) 2026-08-22 16:03:23 +00:00
overtrue b6f135f0fe fix(ecstore): drop removed decommission test import 2026-08-22 23:28:14 +08:00
overtrue b07f0a92a8 merge: sync main into ILM decommission fix 2026-08-22 22:55:14 +08:00
cxymds 2fccfdeabe fix(scanner): fence timed out scan cycles (#6352)
* fix(scanner): fence timed out scan cycles

* fix(scanner): cancel scan workers with cycle scope

* fix(scanner): reject persisted timer overflow

* fix(scanner): reject terminal leadership epochs

* fix(scanner): reject trailing cycle state bytes
2026-08-22 14:38:18 +00:00
Zhengchao An 9815694301 fix(ecstore): supervise decommission worker cleanup (#6372) 2026-08-22 14:33:24 +00:00
houseme 0e79106c2f fix(storageclass): use div_ceil for inline threshold to match shard size calc (#6390)
The inline_block threshold used floor division (DEFAULT_INLINE_OBJECT_BUDGET
/ data_shards) while shard_file_size uses ceiling division (div_ceil). For
EC 12:4 with 256KiB objects, this caused a 1-byte discrepancy:
- inline_block = 262144 / 12 = 21845 (floor)
- shard_file_size = 262144.div_ceil(12) = 21846 (ceil)
- should_inline(21846, 12, false) = false (wrong!)

Fix by using div_ceil for the inline_block calculation, so both sides
use the same rounding and the inline path is correctly triggered.

Co-authored-by: heihutu <heihutu@gmail.com>
2026-08-22 22:05:20 +08:00
Zhengchao An 12a9e654b5 refactor(data-usage): rename ReplicationStats to ReplicationTargetUsage (#6345)
* refactor(data-usage): ReplicationStats -> ReplicationTargetUsage

Rename the data-usage crate's ReplicationStats to ReplicationTargetUsage.
Serde field names are byte-identical (only the Rust type name changed;
field identifiers that rmp encodes are untouched). An rmp round-trip test
guards against future drift.

Scanner test imports updated to match.

* style: cargo fmt
2026-08-22 13:56:58 +00:00
Zhengchao An 6b5e0feef6 test(e2e): wait for heal peers after node rejoin (#6359) 2026-08-22 13:52:08 +00:00
cxymds da90d02c15 test(ecstore): cover suspended-owner heal semantics (#6348)
* test(ecstore): cover suspended-owner heal semantics

* test(heal): cover suspended owner production path
2026-08-22 20:45:03 +08:00
唐小鸭 a930152d5a fix(admin): expose per-target disableProxy through remote target admin API (#6376)
The read-proxy selector already honors a target's disable_proxy flag
(PR #6172), but the admin API still rejected the field, so the only way
to set it was importing a MinIO-written bucket-targets.json.

- move disableProxy from REMOTE_TARGET_UNSUPPORTED_FIELDS to
  REMOTE_TARGET_WRITABLE_FIELDS (set-remote-target create accepts it)
- add TargetUpdateOp::Proxy so set-remote-target?update=true&proxy=true
  overlays only the proxy group (MinIO TargetUpdateType parity)
- bump REMOTE_TARGET_CAPABILITY_CONTRACT_VERSION 1 -> 2 and update the
  runtime capability pin tests
- keep edge/edgeSyncBeforeExpiry rejected (no implementation behind them)
- pin that a published TargetClient carries disable_proxy, the field the
  proxy-target selector consults

Refs rustfs/backlog#1950
2026-08-22 20:43:52 +08:00
cxymds f9d45e41e1 fix(quota): account compressed deletes by committed size (#6365) 2026-08-22 11:39:01 +00:00
overtrue 34a52a1e7d fix(ecstore): restore decommission test imports 2026-08-22 19:13:21 +08:00
cxymds 04e1ea227a fix(scanner): isolate corrupt cycle state (#6354)
* fix(scanner): isolate corrupt cycle state

* fix(scanner): preserve newer state during recovery reset

* fix(scanner): fence recovery reset state

* fix(scanner): reject terminal recovery epochs

* fix(scanner): reject trailing cycle state bytes

* fix(scanner): reject terminal leadership epochs

* fix(scanner): retain recovery wake notifications

* fix(scanner): recover from oversized markers
2026-08-22 11:11:48 +00:00
overtrue fe2d166cd3 fix(ecstore): distinguish active ILM target cleanup 2026-08-22 18:43:05 +08:00
overtrue c687f95260 fix(ecstore): preserve active ILM source journals 2026-08-22 18:43:05 +08:00
overtrue fe66aa509e test(ecstore): isolate durable ILM scenario stack 2026-08-22 18:43:05 +08:00
overtrue 918686bd4e test(ecstore): compile multi-source ILM recovery 2026-08-22 18:43:05 +08:00
overtrue a9cbf82943 test(ecstore): serialize multi-source ILM recovery 2026-08-22 18:43:05 +08:00
overtrue ee7a82ec1f test(ecstore): cover durable ILM recovery boundaries 2026-08-22 18:43:05 +08:00
overtrue 3229ce05e3 fix(ecstore): repair durable ILM receipt recovery 2026-08-22 18:43:04 +08:00
overtrue 7428f3138c fix(ecstore): harden durable ILM cursor receipts 2026-08-22 18:43:04 +08:00
overtrue cf0a78ce11 fix(ecstore): avoid terminal receipt shadowing 2026-08-22 18:43:04 +08:00
overtrue 7bda7f4825 fix(ecstore): re-export durable ILM checkpoint 2026-08-22 18:43:04 +08:00
overtrue 20a8c443af fix(ecstore): anchor decommission ILM receipts 2026-08-22 18:43:04 +08:00
overtrue 948ee1fc52 fix(ecstore): close ILM receipt recovery gaps 2026-08-22 18:43:04 +08:00
overtrue ecd779b6f3 fix(ecstore): track ILM recovery across decommission 2026-08-22 18:43:04 +08:00
overtrue 619fae2512 fix(ecstore): verify ILM metadata before decommission 2026-08-22 18:43:04 +08:00
overtrue 4f9d22bda7 fix(ecstore): migrate ILM metadata during decommission 2026-08-22 18:43:04 +08:00
Zhengchao An 1a3be70d98 fix(ecstore): preserve remote delete error types (#6371) 2026-08-22 17:07:02 +08:00