• Master
  • Supports snowflake file key generation. #1945
• Filer
  • MySQL store ensures name column is case sensitive #1931
• Java Client
  • Support running on windows
• Minor
  • Master avoid possible nil when nodes are disconnected #1928
  • Mount releases resources when Forget() is called.
  • Shell: volume.check.disk adds optional slow mode #1930

• Shell
  • Add volume.check.disk to fix inconsistency for replicated volumes #1923
• S3
  • CopyObject endpoint support copying object to itself #1922
  • dir with slash key does not exist #1917

• FUSE Mount
  • regression: filer.path option results in empty mount in 2.32 #1906
• S3
  • count correction for the number of files in the dir #1904
• Minor
  • remove fasthttp package

• FUSE Mount
  • Resolve memory leak by adding BatchForget in fuse library. The leak was introduced in version 2.23.
• Shell
  • Fix a problem with volumeServer.evacuate and volume.balance #1534
  • add s3.clean.uploads to clean up stale multipart uploads #1855
• Minor
  • S3 escape special characters in file names #1884
  • Filer leveldb, rocksdb metadata store automatically create metadata folers #1901
  • Filer mysql2, postgres2 trigger actions on bucket creation and deletion #1877
  • -ip.bind default to empty

Fixing a problem that .vif file for each volume is not generating. It can cause failures when moving volumes during rebalancing or fix replication. #1878

• Filer
  • Adds weed filer.meta.backup command to backup metadata to a different filer store.
• Cloud Tier
  • Upload to non-AWS S3 gateways
• S3
  • ListBuckets show accessible buckets #1837
  • MultipleObjectDelete also deletes folders #1857
• Volume Server
  • Fix EC shard count logic #1863
  • Add experimental TCP support
• In general
  • Support IPv6
  • Misc fixes for nil objects

• Add filer.backup command
  • Avoids external message queue. Supports S3, GCP, Azure, B2, local disk, etc.
  • Adds incremental backup mode for daily new or updated files.
• filer.tier.move
  • avoid repeated move for replicated volumes #1792
• S3
  • return http status code 200 for empty files #1830 #1831
• Java Client
  • ByteBuffer.clear supports higher JDK versions #1836
• command line options
  • Use -options for saved CLI options https://github.com/chrislusf/seaweedfs/wiki/Use-Config-File-for-Command-Line-Options

This is a minor release. Adding back weed scaffold -config=xxx

• Tiered Storage
  • Support custom tags for tiers.
  • Add volume.tier.move to move data from one tier to the other.
• Minor fixes
  • filer: avoid encryption and compression at the same time #1828
  • volume: avoid fixed vacuum timeout for large volumes

• Support command line arguments in a configuration file #1823
• Filer
  • Fix postgres2 mysql2 avoid repeatedly creating tables
  • Fix fs.configure should try to read from entry content #1792
• Volume
  • Heartbeat should fail fast #1820
• Minor
  • filer: do not print password on error #1809
  • webdav: add replication setting #1817

S3

• Fix a critical bug with S3 in 2.24, 2.25 releases #1806 Critical - S3 Delete Multiple Objects, delete the bucket
• Tiering
  • Support SSD volumes https://github.com/chrislusf/seaweedfs/wiki/Tiered-Storage
• Java
  • Refactored SeaweedFS Java client https://github.com/chrislusf/seaweedfs/wiki/SeaweedFS-Java-Client
• Kubernetes
  • Updated Helm charts #1789 #1790
• FUSE mount
  • Fix with correct filer path #1799
  • Ensure proper release #1803
• Minor
  • Fail fast if idx files are missing #1796
  • Replication has a few changes with RabbitMQ #1801
  • Add http max idle connections per host #1802
  • Performance increase with fasthttp client
  • Mount remove unnecsary flush on release

• WebDAV
  • Can start together with filer or server as weed filer -webdav or weed server -webdav.
• S3
  • DeleteMultipleObjects efficiently clean up leftover empty folders #1772
  • Return 204 when deleting nonexistent files #1776
• Replication
  • RabbitMQ adds delayed retries with dead letter exchange #1773
• Filer
  • Fix regression with ElasticSearch filer store #1774
• Java
  • Adds official Java client API with examples, with user friendly SeaweedInputStream and SeaweedOutputStream.
• Hadoop compatible file system
  • Adds optional fs.seaweed.volume.server.access to access volume servers via their publicUrl settings, or use filer as a proxy.
  • Enabled flush() function.

• Filer

  • Fix filer store leveldb and hbase miss files when listing large directories more than 1024 entries #1768
  • Fix filer store etcd listing bugs #1767

• FUSE Mount

  • Invalidate kernel cache properly #1752
  • Add volumeServerAccess modes for publicUrl and filerProxy
  • Report error when flush()

• S3

  • Optimized DeleteMultipleObjectsHandler

• Replicate

  • Add local and local_incremental to backup to local file systems #1759 #1761 #1762

• Shell

  • volume.fix.replication adds collectionPattern with wildcard characters #1758

• Minor

  • Fix concurrent map reads
  • FUSE Mount on FreeBSD skip daemon timeout
  • Avoid possible nil reader #1754

• Filer Store
  • Add mysql2, postgres2 for dedicated tables for each bucket. This ensures instant bucket deletion even for large buckets.
• Filer
  • Support appending to a file with "op=append" URL parameter.
  • Fix http status 206 setting #1753
• Fuse Mount
  • Use DIRECT_IO on linux to avoid page cache and ensure write-once-read-many with multiple weed mount instances. #1752
  • Limit background requests and congestion threshold
• Filer meta
  • Watch meta data changes in JSON format by weed filer.meta.tail.
• Filer Sync
  • weed filer.sync can optionally use filer as a proxy to access volume servers, avoid network setup to expose all volume servers.
• S3
  • Fix for listing objects if more than 1000 in the folder

• Filer
  • Small values are saved to filer store only, saving one network hop to volume server.
  • Fix bug to ensure seamless meta data replication under stress loads.
  • Add leveldb3 to have dedicated filer store for each bucket, for instant large bucket deletion.
  • Avoid http status 206 when error happens for range requests.
• Mount
  • Fix a data read error due to a race condition when write and then read back quickly.
• Filer meta tail
  • Add feature to submit metadata updates to ElasticSearch.
• S3
  • Bug fix: #1741 key count during list objects
• Minor fixes
  • Fix viper concurrent modification when starting servers.
  • Filer store: nil entry error if not found

• Mount
  • Default to 128 concurrent writers to avoid OOM
• Filer
  • RocksDB adds native support for TTL
  • Upgrade redis module
• Cloud Tiering
  • Fix s3 upload "use of closed network connection" #1734
• Add weed filer.cat command
  • Read files directly from volume servers, without proxying through filer
• Server
  • Add option to disable master

New

• Filer
  • Add RockDB as filer store

Bug fixes

• S3
  • Fix prefix listing #1722
  • Respect Content-Type on PUT #1720
• Shell
  • Support shell.toml to switch cluster easily #1664
• Mount
  • Fix weed mount with TLS #1714
• Master
  • Fix nil snapshot #1718

• S3
  • Fix a regression causing nil #1707
  • Add -allowEmptyFolder option to disable auto removing empty folders.
• Filer
  • Add namePattern to search files by name within one folder

• Filer
  • Redis, Cassandra can support super large directories
  • Add HBase for filer store
  • Make path-specific stores portable by trimming out common prefixes
  • Add -defaultStoreDir so that filer.toml can be skipped #1659
• S3
  • Fix a bug causing loop with empty folders #1701
  • Fix bucket-specific Admin permissions

• Volume Server
  • Avoid repeated loading the same volume. #1682
• Filer
  • Support path-specific filer store. (EXPERIMENTAL in 2.16, there are incompatible changes in the git tip)The meta data for one branch of the directory tree can be stored on a separate filer store, for better efficiency and scalability.
  • Fix Mysql/Postgres error handling for non-english languages.
  • Return http status 499 for client closed connection. #1684
• Minor
  • Expose master UI when http APIs are disabled. #1650
  • Detect IP address better. #1671
  • Prevent mistakenly delete empty collection. #1677
  • filer.copy errors a little bit more obviously. #1678

• Shell
  • fix fs.configure for path-specific storage configuration, e.g., Replication, TTL, Collection.
  • add s3.configure to update the s3 identities without restart
• FUSE mount
  • Avoid memory leaking
• S3
  • move "delete-directory-if-empty" operation to read time
  • Fix listing with sub folders #1656
  • Avoid cross-collection file rename
• Filer
  • Paginate through large folders
  • Change to /etc/seaweedfs/filer.conf on filer for path-specific configuration.