What seaweedfs/seaweedfs shipped

SeaweedFS 4.53 is out. Volume encryption now follows the filer's single switch across s3, vector, worker and CLI. S3 Vectors writes tables straight to filers and volume servers. Versioned deletes, EC quarantine and writeback retries got fixes.

SeaweedFS 4.53 is released. In this week's changes: volume encryption now follows the filer's -encryptVolumeData switch across weed s3, weed-vector, weed-worker and weed-vector-cli, so one setting governs it everywhere. S3 Vectors gained help text fixes and now writes Lance table files directly to filers and volume servers rather than through the S3 port. Correctness work landed for versioned S3 deletes, including minting a PUT's version id at commit and not treating filer errors as successful deletes. Quarantined EC shards no longer serve reads, and writeback chunk uploads keep retrying through longer volume outages. Single-master restarts also stop polling quorum every millisecond.

This week: raft PreVote keeps a returning master from deposing the leader, gRPC TLS fails closed, S3 bucket-scoped Admin grants auto-create buckets, filer.backup forwards user metadata and gains -s3.flattenVersions, and vector queries cache Lance reads.

Changes this week in SeaweedFS: - raft bumped to v1.2.3 with PreVote: a master returning from a partition no longer deposes the healthy leader. - gRPC TLS fails closed on misconfiguration and verifies the dialed server identity. - S3 gateway drops retired volume servers from its location cache. - Vector engine caches write-once Lance file reads across requests. - Ranged GET on the Go volume server avoids reading whole needles of 1MB or less. - filer.backup forwards user metadata and gains -s3.flattenVersions. - POST Object uploads auto-create missing buckets; bucket-scoped Admin grants can auto-create matching buckets. - Helm chart adds commonLabels and namespaceOverride.

SeaweedFS 4.51 is out: S3 GET and HEAD no longer echo stored tags and internal attributes, read-only admin sessions stay out of the filer /etc tree, and anonymous grants on table catalogs are scoped. Also this week: delete marker responses, a shorter first-write hold, and no extr

SeaweedFS 4.51 is out. S3 GET and HEAD responses no longer echo stored tags or internal attributes, read-only admin sessions are kept out of the filer /etc tree, and the anonymous identity's grant on the table catalogs is now scoped. Other changes this week: GET and HEAD of a delete marker answer as S3 does instead of a bare 404; the first write after a fresh topology id no longer waits out the warmup window; commands bound to a specific -ip.bind address no longer open an extra localhost listener; S3 POST policy uploads authorize through the shared object check; ListObjectsV2 continuation tokens round-trip keys XML cannot carry; and bucket tag sets accept up to 50 tags.

This week: PUT with extract=zip expands archives into files on the filer and S3, the daily lifecycle pass shares one bucket walk across shards, S3 listing returns children of a StartAfter directory, and bucket delete counts stored versions.

Shipped this week in SeaweedFS: zip archives can now be expanded on upload, both through the filer (extract=zip or X-Seaweedfs-Extract) and through S3 PutObject, with each member written as its own file. The daily lifecycle pass now shares a single bucket walk across shards instead of listing the tree once per shard, and replay no longer stalls when a bucket has rules longer than the retention window. S3 listing now returns the children of a directory named by a StartAfter marker, bucket delete counts stored versions as contents, and storage rules with a rack or data-center constraint are honored rather than falling back. Vacuum commits are bounded per volume server, SFTP rename enforces delete permission on an existing target, and the admin Create Bucket button sends one request instead o

This week: filer cache contention fixed for small S3 reads, stable ETags across remote cache fills, mount failover to the next filer, synced-second rewrites no longer dropped, identity revocation on config reload, and Azure Key Vault SSE-KMS round trips fixed.

Changes this week in SeaweedFS: - Small concurrent S3 GETs no longer serialize on the filer reader cache mutex. - Remote-mounted objects keep a stable ETag across cache fills, so ranged downloads stop failing. - Mounts with several filers fail writes over to the next reachable one instead of hanging. - A filer announces to the master only once its gRPC port serves, and leaves before it stops. - Filer writes fail when the existing-entry lookup fails rather than replacing an entry without its chunks. - Filer remote sync no longer drops rewrites landing in the same second as the previous upload. - Removing an identity from the config and reloading now revokes its access keys. - Azure Key Vault SSE-KMS encrypt and decrypt round trips work, and volume data encryption is applied on copy.

Seaweedfs merged 40 PRs this week: S3 ranged-read speedup (7x on some workloads), persistent deletion ledger to prevent orphaned chunks, metadata sync recovery improvements, and OIDC role management via gRPC.

Seaweedfs merged 40 improvements this week addressing reliability and performance. Notable: S3 ranged GETs now avoid full-chunk downloads (7x faster on range-heavy workloads); filer deletion queue persists to disk to prevent orphaned chunks on restart; metadata sync resubscribes correctly on target failures; S3 server tracks live filer topology; and the IAM service now manages OIDC providers and roles. These changes reduce data loss risk, improve query latency, and strengthen multi-filer deployments.

Weekly fixes: S3 authorization gaps closed, volume locks held shorter, empty volumes now reclaimed, corrupt needles rejected, lock ring recovery fixed, and REST API docs restored.

This week in SeaweedFS: fixed S3 authorization bypasses where prefix parameters leaked between actions, shortened volume server lock holds to prevent heartbeat stalls and memory exhaustion, enabled vacuum to reclaim empty volume slots at capacity, rejected corrupt needle delivery, fixed master lock ring poisoning after failover, and restored the HTTP REST API reference documentation.

4.47.1 fixes: EC BatchDelete validation, negative Size rejection preventing node brick, tomb­stone scrubbing, corrupt needle loops, multipart upload atomicity, and orphaned chunk safety.

Seaweedfs 4.47.1 addresses critical data integrity issues: EC BatchDelete now validates cookies preventing bypass attacks, negative Size values are rejected before poisoning store locks, corrupt needle headers no longer cause infinite loops, and S3 multipart uploads complete atomically. Erasure coding validation at gRPC boundaries prevents ShardId truncation attacks. S3 orphaned chunk deletion waits for entry confirmation. Volume scrubbing now processes tombstones correctly.