57403525ee
refactor: move heal and namespace contracts ( #3560 )
2026-06-18 11:00:21 +08:00
339a192a09
fix(ecstore): preserve bucket visibility during rebalance ( #3418 )
...
Co-authored-by: houseme <housemecn@gmail.com >
2026-06-14 03:37:54 +08:00
0cdcd1eb7b
feat(table-catalog): internalize catalog backing paths ( #3295 )
...
* feat(table-catalog): internalize catalog backing paths
* fix(table-catalog): clean internal catalog on bucket delete
---------
Co-authored-by: Henry Guo <marshawcoco@users.noreply.github.com >
2026-06-09 12:01:25 +00:00
78df276d25
fix: return 503 on lock contention instead of 500 ( #3274 )
...
* 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 >
2026-06-08 02:20:21 +00:00
d6120f5788
refactor(ecstore): migrate mutable globals into ECStore struct fields ( #3214 )
...
* refactor(ecstore): migrate mutable globals into ECStore struct fields
Phase 1 of global singleton consolidation. Move mutable globals from
lazy_static into ECStore struct fields as the first step toward
dependency injection and multi-instance support.
New ECStore fields:
- is_erasure, is_dist_erasure, is_erasure_sd (erasure type flags)
- local_disk_map, local_disk_id_map, local_disk_set_drives
- root_disk_threshold
- tier_config_mgr, event_notifier, bucket_monitor
New accessor methods:
- is_erasure(), is_dist_erasure(), is_erasure_sd()
- update_erasure_type()
- tier_config_mgr(), event_notifier(), bucket_monitor()
Global functions in global.rs preserved for backward compatibility.
1151 ecstore tests pass.
* fix: address PR #3214 review comments
- Sync ECStore fields from globals after init()
- Enforce DistErasure => is_erasure invariant in update_erasure_type()
- Change bucket_monitor from Option to OnceLock for deferred initialization
- Restrict TypeLocalDiskSetDrives to pub(crate)
- 1151 tests pass
* fix: address PR #3150 review comments
- Restore get_host_addr as best-effort wrapper (String return type)
- Replace bare expect("err") with descriptive messages in unsigned trailer
- Simplify aws-chunked header construction
- 20 signer + 96 io-core tests pass
* fix(ecstore): format store imports
* fix(ecstore): keep migrated accessors in sync
* fix(signer): preserve host fallback and unsigned trailer errors
* fix(ecstore): defer migrated global accessors
* fix(signer): box unsigned trailer signing errors
* fix(ecstore): avoid lock awaits during sync
* fix(ecstore): narrow phase one globals
---------
Co-authored-by: houseme <housemecn@gmail.com >
2026-06-04 17:27:21 +00:00
be98c1f86a
feat(table): add catalog boundary primitives ( #3173 )
...
Co-authored-by: Henry Guo <marshawcoco@users.noreply.github.com >
2026-06-03 13:16:59 +00:00
97dd825468
fix(ecstore): optimize ObjectInfo clone and fix critical TODOs ( #3149 )
...
* fix(ecstore): optimize ObjectInfo clone and fix critical TODOs
## ObjectInfo hot-path clone optimization (issue #653 item 2)
- Wrap user_defined (HashMap), user_tags (String), parts (Vec) in Arc
- Clone cost reduced from ~20 heap allocations to O(1) ref count bump
- Updated 11 downstream access sites with explicit deref where needed
## Critical TODO/FIXME fixes (issue #653 item 7)
- rpc/peer_s3_client.rs: descriptive error message for empty peer response
- set_disk/write.rs: concurrent rollback deletes via tokio::spawn + join_all
- store_list_objects.rs: resolved FIXME with explanation + 4 regression tests
- store/bucket.rs: namespace write locks for make_bucket and delete_bucket
Skipped TODOs (too risky without broader context):
- multipart.rs:30 nslock — causes lock timeout in existing tests
- bucket.rs:88 cached list_bucket — needs cache invalidation strategy
- bucket.rs:149 replication delete — needs replication subsystem integration
1134 tests pass (1 flaky under parallel execution, passes individually).
* fix: address PR #3149 review comments
- Preserve NamespaceLockQuorumUnavailable variant in bucket create/delete
- Fix cancellation test to actually test cancellation (keep senders open)
- Update ObjectInfo consumers in app/ for Arc-backed fields
- Update bucket_usecase.rs user_tags to Arc<String>
* fix: update ObjectInfo Arc consumers
* fix: address ecstore merge review comments
* fix: cancel blocked merge output sends
* fix(ecstore): satisfy ObjectInfo clone clippy
2026-06-01 00:44:58 +00:00
236142a682
fix(ecstore): repair lifecycle transition and restore flows ( #2240 )
...
Co-authored-by: heihutu <heihutu@gmail.com >
Co-authored-by: 安正超 <anzhengchao@gmail.com >
Co-authored-by: houseme <housemecn@gmail.com >
Co-authored-by: loverustfs <hello@rustfs.com >
Co-authored-by: weisd <im@weisd.in >
2026-03-23 12:29:13 +08:00
c6715259b1
fix(ecstore): keep bucket created time when metadata is epoch ( #2220 )
...
Co-authored-by: momoda693 <momoda693@gmail.com >
Co-authored-by: houseme <housemecn@gmail.com >
2026-03-19 01:02:34 +08:00
095b77795d
refactor(ecstore): split store.rs into store submodules ( #1942 )
2026-02-25 06:35:27 +08:00
安正超