Commit Graph

4951 Commits

Author SHA1 Message Date
马登山 29c2f247c3 Merge remote-tracking branch 'origin/main' into cxymds/fix-transition-identity-flake 2026-07-29 15:09:04 +08:00
cxymds d9efd6b853 feat(ecstore): add remote snapshot lease RPCs (#5389)
* feat(ecstore): add local snapshot leases

* feat(ecstore): add remote snapshot lease RPCs

* fix(rpc): keep snapshot lease checks CI-compatible
2026-07-29 15:06:18 +08:00
cxymds 2801b2500d fix(tiering): gate remote version state writes (#5382)
* feat(tiering): model provider version capabilities

* feat(tiering): persist opaque remote versions

* fix(tiering): gate remote version state safely

* fix(tiering): gate remote version state writes

* fix(tiering): preserve remote version state on delete

* fix(tiering): accept unversioned transition responses

* fix(tiering): replay exact cleanup journals

* test(tiering): pin empty exact cleanup guard

* test(tiering): accept strict missing journal errors

* test(tiering): exercise free-version identity guard

* test(tiering): reach destination identity guard

* test(tiering): persist version identity drift

* test(tiering): bind version drift fixture

* test(config): keep fleet gate tests after constants
2026-07-29 15:06:08 +08:00
cxymds 02f4dbeb68 fix(heal): fail closed on ambiguous metadata rescue (#5361)
* fix(heal): fail closed on ambiguous metadata rescue

* fix(heal): preserve uncertain dangling state

* test(heal): satisfy strict clippy checks

* fix(heal): retain explicit version metadata recovery

* fix(heal): preserve metadata rescue diagnostics

---------

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 15:06:01 +08:00
cxymds f7757e6437 test(ecstore): rendezvous concurrent resend commits (#5411)
* test(ecstore): rendezvous concurrent resend commits

* test(ecstore): satisfy multipart barrier lint
2026-07-29 15:05:53 +08:00
cxymds 1cb1b02b08 fix(ecstore): prevent deleted bucket recreation (#5380)
* fix(ecstore): prevent deleted bucket recreation

* fix(ecstore): reject empty metadata bucket names

* fix(ecstore): avoid recursive bucket metadata lookup

* fix(ecstore): bound lazy metadata future stack use

* test(ecstore): create bucket before metadata reload

---------

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 14:49:18 +08:00
Zhengchao An 3fe74a5019 fix(swift): merge account and container metadata POSTs (#5414) 2026-07-29 06:36:07 +00:00
Zhengchao An 451cbc099b fix(tier): keep converging peers that reject a config reload (#5412) 2026-07-29 14:24:26 +08:00
cxymds b683bf236e Merge branch 'main' into cxymds/fix-transition-identity-flake 2026-07-29 14:00:26 +08:00
cxymds cb62079ba6 test(ilm): make active cancellation deterministic (#5403)
* ci: preserve timeout evidence for workspace tests

* test(ilm): make active cancellation deterministic

* ci: allow cold doctest compilation to finish

---------

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 13:53:59 +08:00
Zhengchao An d39ffdb1cd fix(filemeta): canonicalize version order after insert (#5413) 2026-07-29 05:49:00 +00:00
马登山 9aa5a7d956 Merge remote-tracking branch 'origin/cxymds/fix-rio-resend-lock-timeout' into cxymds/fix-transition-identity-flake 2026-07-29 13:38:06 +08:00
马登山 0dcf929ebc Merge remote-tracking branch 'origin/main' into cxymds/fix-transition-identity-flake
# Conflicts:
#	crates/ecstore/src/set_disk/ops/object.rs
2026-07-29 13:37:51 +08:00
马登山 1ae87c5990 test(ecstore): satisfy multipart barrier lint 2026-07-29 13:36:39 +08:00
cxymds 7ac419fe39 Merge branch 'main' into cxymds/fix-rio-resend-lock-timeout 2026-07-29 13:23:43 +08:00
Zhengchao An 7d698abc1f ci(checksums): inherit workspace lint policy (#5415) 2026-07-29 12:53:45 +08:00
hector a1a65ad65d fix(action): change quay.io image repository name (#5417) 2026-07-29 12:53:26 +08:00
cxymds 358af6a8de ci: preserve timeout evidence for workspace tests (#5402)
* ci: preserve timeout evidence for workspace tests

* ci: allow cold doctest compilation to finish

* ci: allow cold nextest compilation to finish

* ci: allow cold nextest compilation to finish

---------

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 04:49:03 +00:00
Zhengchao An 90d1a15d13 fix(ecstore): classify peer RPC failures by gRPC status code (#5400) 2026-07-29 11:35:09 +08:00
马登山 a02ca95093 test(ecstore): rendezvous concurrent resend commits 2026-07-29 11:28:06 +08:00
cxymds 2423ba8e3f fix(tiering): base restore expiry on completion (#5366)
* fix(tiering): base restore expiry on completion

* fix(tiering): import restore metadata types

* fix(restore): resolve metadata finalization build errors

* test(ecstore): import restore expiry helper
2026-07-29 02:55:00 +00:00
cxymds 294c79c156 fix(tiering): gate remote version state safely (#5374)
* feat(tiering): model provider version capabilities

* feat(tiering): persist opaque remote versions

* fix(tiering): gate remote version state safely

* fix(tiering): preserve remote version state on delete

* fix(tiering): accept unversioned transition responses

* fix(tiering): replay exact cleanup journals

* test(tiering): pin empty exact cleanup guard

* test(tiering): accept strict missing journal errors

* test(tiering): exercise free-version identity guard

* test(tiering): reach destination identity guard

* test(tiering): persist version identity drift

* test(tiering): bind version drift fixture

---------

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 02:43:28 +00:00
cxymds 3d80578abd feat(ecstore): add local data-dir snapshot leases (#5388)
feat(ecstore): add local snapshot leases

Co-authored-by: houseme <housemecn@gmail.com>
2026-07-29 02:42:35 +00:00
Zhengchao An 957080bea5 fix(swift): persist container and account metadata writes (#5398)
Swift container and account metadata handlers cloned the cached
BucketMetadata, set the tagging fields, and called set_bucket_metadata,
which only updates the in-memory cache map. Nothing reached
.metadata.bin, so every Swift metadata POST was lost on restart and
silently overwritten by the next disk-truth reload (a peer
LoadBucketMetadata notification or the 15-minute refresh loop) — while
the client had already been told 2xx.

Route these writes through a new metadata_sys::update_config_with: a
read-modify-write that loads the on-disk metadata and persists the
result under the same write guard metadata_sys::update uses, so the
rewrite merges against disk truth instead of a possibly stale cache and
cannot clobber a concurrent update to another config file. Peers are
notified afterwards, matching the S3 config handlers.

Persisting these writes required hardening the paths that now produce
durable state:

- Account metadata writes validate account ownership. This metadata
  holds the account's TempURL signing key, so an unauthenticated write
  for someone else's account would have become a durable, cluster-wide
  takeover of that account's pre-signed URLs. Reads stay open because
  TempURL signature validation runs before credentials exist.
- disable_versioning verifies the container exists. Without it the
  metadata loader's "no metadata on disk" default would be persisted,
  creating an orphan metadata file and caching a fabricated default as
  authoritative.
- Container and account metadata are size- and count-limited, reusing
  the Swift limits object metadata already enforces; these tags land in
  the bucket metadata file that every later config write rewrites whole.
- A rewrite refuses to run when the persisted tagging config is
  unreadable, instead of merging onto an empty set and wiping the
  container ACL and versioning tags. It reports 409 naming the remedy.
- Storage errors are logged in full and reported generically, since they
  now carry real disk and quorum detail.

The tagging arm of BucketMetadata::update_config also clears the parsed
config, as the lifecycle arm does: parse_all_configs skips empty XML
rather than clearing, so a cleared config kept serving the old tags.
Tagging is serialized with the S3 XML serializer the loader can parse
back, not quick_xml, whose output was never round-trippable.
2026-07-29 02:31:11 +00:00
houseme c6d42a5355 Merge branch 'main' into cxymds/fix-transition-identity-flake 2026-07-29 09:58:50 +08:00
cxymds c1538cf1c3 fix(multipart): serialize complete and abort (#5356)
* fix(multipart): serialize complete and abort

* test(multipart): order abort-first finalization

* fix(multipart): enforce quorum staging cleanup

* fix(multipart): remove stale mutable binding
2026-07-29 01:49:19 +00:00
houseme 5af56cbb02 test(ci): stabilize lifecycle timeout coverage (#5404)
Co-authored-by: heihutu <heihutu@gmail.com>
2026-07-29 01:37:29 +00:00
houseme f99956eade test(lifecycle): classify mixed rollout harness results (#5384)
* test(lifecycle): classify mixed rollout harness results

Classify Docker #1508 evidence as strict, baseline, blocked, or failed so tiered-storage baseline runs cannot be mistaken for strict mixed-version rollout closure evidence.

Co-Authored-By: heihutu <heihutu@gmail.com>

* test: classify Docker manual transition preemption

Co-Authored-By: heihutu <heihutu@gmail.com>

---------

Co-authored-by: heihutu <heihutu@gmail.com>
2026-07-29 01:21:23 +00:00
houseme 775279b6fd fix(notify): defer disabled bootstrap storage refresh (#5373)
Co-authored-by: heihutu <heihutu@gmail.com>
2026-07-28 17:39:22 +00:00
马登山 4826c0207e test(ecstore): remove invalid transition version drift case 2026-07-28 20:38:18 +08:00
cxymds eb755e2b97 fix(auth): compare sensitive tokens in constant time (#5371)
* fix(auth): compare sensitive tokens in constant time

* test(auth): scope constant-time source guard

* test(auth): ignore test source in comparison guard

* chore(deps): use constant-time s3s authentication
2026-07-28 17:56:22 +08:00
cxymds 7f146fc5de feat(tiering): model provider version capabilities (#5372) 2026-07-28 17:42:08 +08:00
cxymds 5426237a49 fix(filemeta): preserve canonical version order (#5353) 2026-07-28 17:38:59 +08:00
cxymds d7afa4e38e fix(heal): report partial rename failures (#5355)
* fix(heal): report partial rename failures

* fix(log-analyzer): track partial heal rename failures
2026-07-28 17:37:02 +08:00
cxymds a3f5a8eaf9 fix(ecstore): fence object tagging updates (#5360)
* fix(ecstore): fence object tagging updates

* test(ecstore): stabilize tagging lock regressions

* test(ecstore): restore setup type on current-thread runtime
2026-07-28 17:36:00 +08:00
cxymds 547c678eed fix(filemeta): reject positive size without parts (#5354)
* fix(filemeta): reject positive size without parts

* test(ecstore): keep optimized read fixture valid

* test(ecstore): keep listing fixtures valid
2026-07-28 17:35:53 +08:00
cxymds df945b275a fix(lifecycle): recover saturated scanner tasks (#5369) 2026-07-28 17:30:52 +08:00
cxymds 2e78a49c95 fix(lifecycle): reject invalid modification times (#5370) 2026-07-28 17:23:21 +08:00
cxymds 7ad0e726db fix(ecstore): use set read quorum for version reads (#5365)
* fix(ecstore): use set read quorum for version reads

* fix(ecstore): convert optimized read batch errors
2026-07-28 17:23:15 +08:00
cxymds 45c3386b68 fix(bitrot): reject trailing shard data (#5358)
* fix(bitrot): reject trailing shard data

* test(ecstore): align bitrot disk fixture type
2026-07-28 17:23:08 +08:00
cxymds 7af92b4f54 fix(ecstore): handle pre-epoch disk health clocks (#5383) 2026-07-28 17:22:26 +08:00
cxymds daa627ee0e fix(auth): enforce presigned URL expiry limit (#5368)
* fix(auth): enforce presigned URL expiry limit

* ci(deny): allow presigned expiry s3s fork

* chore(deps): update presigned expiry validation
2026-07-28 17:11:30 +08:00
cxymds 5c3d3a8220 fix(ecstore): handle mutex poisoning explicitly (#5379)
* fix(ecstore): handle mutex poisoning explicitly

* test(ecstore): satisfy poison assertion lint
2026-07-28 17:09:09 +08:00
cxymds 6bc5fc77b5 fix(notify): emit noop event for missing deletes (#5381) 2026-07-28 17:05:10 +08:00
cxymds e822fc1552 fix(swift): enforce cumulative container quotas (#5378) 2026-07-28 17:05:04 +08:00
cxymds 2f6115e058 fix(swift): enforce TempURL IP restrictions (#5377) 2026-07-28 17:04:58 +08:00
cxymds 882e1a71b1 fix(tiering): abort failed multipart restores (#5367)
* fix(tiering): abort failed multipart restores

* fix(tiering): compile restore abort regressions

* test(tiering): assert restored multipart metadata

* test(tiering): parse completed restore status
2026-07-28 17:04:52 +08:00
cxymds b432f31c2c fix(multipart): preserve retried parts on quorum failure (#5363)
* fix(multipart): preserve retried parts on quorum failure

* style(multipart): format transaction rollback

* fix(proto): regenerate multipart transaction RPCs

* fix(multipart): import rollback marker constant

* fix(multipart): export transaction action

* fix: import multipart transaction test requests
2026-07-28 17:04:46 +08:00
cxymds 6e0640444e fix(multipart): list uploads across all sets (#5362) 2026-07-28 17:04:40 +08:00
cxymds 4fb9b0dc7f fix(authz): fail closed on policy load errors (#5359)
* fix(authz): fail closed on policy load errors

* test(authz): cover policy failure precedence

* test: align policy failure expectations

* test: align upload part copy fail-closed expectation
2026-07-28 17:04:35 +08:00