Broker
- Fix incorrect number of read compacted entries #20978
- Fix message loss during topic compaction #20980
- Fix get outdated compactedTopicContext after compactionHorizon has been updated #20984
- Fix compaction subscription delete by inactive subscription check #20983
- Fix incorrect unack msk count when dup ack a message #20990
- Fix ModularLoadManagerImpl always delete active bundle-data #20620
- Use MessageDigest.isEqual when comparing digests #21061
- Fix can't stop phase-two of compaction even though messageId read reaches lastReadId #20988
- Make sure all inflight writes have finished before completion of compaction #21067
- Fix chunked messages will be filtered by duplicating #20948
- Fix potential case cause retention policy not working on topic level #21041
- Make read compacted entries support maxReadSizeBytes limitation #21065
- Remove bundle-data in local metadata store #21078
- Fix write duplicate entries into the compacted ledger after RawReader reconnects #21081
- Cleanup correctly heartbeat bundle ownership when handling broker deletion event #21083
- Avoid splitting one batch message into two entries in StrategicTwoPhaseCompactor #21091
- Fix unsubscribe non-durable subscription error #21099
- Fix isolated group not work problem #21096
- Consider iowait as idle #19110
- Fix deleting topic not delete the related topic policy and schema #21093
- Revert remove duplicate topics name when deleteNamespace #21087
- Improve logs for troubleshooting #21141
- Fix web tls url null cause NPE #21137
- Fix unack count when mixing non batch index and batch index acks #21126
- Backport fix UniformLoadShedder selecet wrong overloadbroker and underloadbroker #21180
- Fix bug caused by optimistic locking #18390
- Disable balancing based on DirectMemory #21168
- Fix replicator leak when removeReplicator in NonPersistentTopic #21205
- Fixed reset for AggregatedNamespaceStats #21225
- Fixed produce and consume when anonymousUserRole enabled #21237
Pulsar IO and Functions
- [FN] Fix ProducerConfig cannot update error #21037
- [FN] Fix the --batch-builder not working error for functions #21023
- [FN] Fix functions_log4j2.xml delete strategy config #21215
- [IO] Fix --retain[-key]-ordering not working error for sink #21060
Client
- Fix perf-producer get OOM with high publish latency #20914
- Fix RawReader hasMessageAvailable returns true when no messages #21032
- Fix consumer can't consume resent chunked messages #21070
- Fix cannot retry chunk messages and send to DLQ #21048
- Fix logging problem in pulsar client #21094
- Avoid ack hole for chunk message #21101
- Fix repeat consume when using n-ack and batched messages #21116
- Fix same producer/consumer use more than one connection per broker #21144
- Merge lookup requests for the same topic #21232
CLI
- [Admin] Fix get topic stats fail if a subscription catch up concurrently #20971
Others
- [META] Fix deadlock in AutoRecovery #21010
- [META] Improve fault tolerance of blocking calls by supporting timeout #21028
- [META] Improve to the ReplicaitonWorker performance by deleting invalid underreplication nodes #21059
- [META] Fix metadata store deadlock due to BookkeeperInternalCallbacks.Processor #21159
- [META] Fix PulsarLedgerUnderreplicationManager notify problem #21161
- [META] Fix pulsar ledger auditor dead lock problem #21181
- [ML] Persist mark deleted ops to ZK if create cursor ledger was failed #20935
- [BK] Improve getIsolationGroup by avoid creating arrayList #20952
- [BK] Fix RocksDB configuration #21157
- [Proxy] Support disabling metrics endpoint #21031
- [Proxy] Fix Proxy 502 gateway error when it is configured with Keystore TLS and admin API is called #21077
- [TXN] Fix the consumer stuck due to deduplicated messages in pending ack state #21177