Releases: apache/celeborn
Releases · apache/celeborn
v0.1.4
- optimize RPC memory usage
- keep one copy of roaring bitmap
- shade commands-lang3 to avoid dependency conflicts
- update RSS dashboard
Full Changelog: v0.1.3...v0.1.4
v0.1.3
What's Changed
- Add map id range read filter. @FMX #607 #660 #662
- Fix worker replied ack at the wrong time when a soft split is trigger… by @waitinfuture in #647
- Suport zstd by @waitinfuture in #652
Full Changelog: v0.1.2...v0.1.3
v0.1.2
This release is recommended and contains various bugfix.
MD5 (rss-0.1.2-bin-release-with-spark3.tgz) = 3bc9bde4773ce9c383c0506a36ede2b1
MD5 (rss-0.1.2-bin-release-with-spark2.tgz) = 2e8b51f8d019441cc50863b74afd1683
What's Changed
- [ISSUE-496][BUG]Ensure all file channels write whole buffer. (#561) by @waitinfuture in #562
- [ISSUE-567] Optimize LifecycleManager.getLatestPartition by @waitinfuture in #571
- [ISSUE-567][FOLLOW-UP] remove entry from latestPartitionLocation in r… by @waitinfuture in #576
- Cherry-pick transport layer refactor from main branch by @waitinfuture in #587
- [ISSUE-588][BUG] Fix memory leak in shuffle read by @waitinfuture in #590
- [ISSUE-591][BUG] Incorrect result when revive and split happen concur… by @waitinfuture in #593
- [IMPROVEMENT] Improve revive or partition split log readability. by @FMX in #603
- [ISSUE-625][BUG] Incorrect result when kill worker while pushMergedData by @waitinfuture in #626
- [ISSUE-281][BUG] Use correct maxDestLength to check if buffer can satisfy compress result By @AngersZhuuuu in #282
- [BUG]Fix slots release when cleanup expired shuffle By @Flyangz in #514
- [ISSUE-413] Fix incorrect totalSpace in disk usage check By @lichaojacobs in #414
- [Issue 253] Fix netty memory leak in TransportResponseHandler By @dxheming in #254
- [BUG] DiskFlusher nextWorkerIndex update incorrectly By @FMX in #295
- [BUG] After trigger pausePushData, new added channel still can push data By @FMX in #376
- fix clean disk failure when the disk is full. fix concurrent issue in PartitionFilesSorter. @FMX
Full Changelog: v0.1.1...v0.1.2
v0.1.1
What's Changed
- [BUG] Fix worker pushserverPort and replicateServerPort conflict by @dxheming in #165
- [BUG]fix an ArithmeticException. by @FMX in #167
- [ISSUE-169][FEATURE] Make app heartbeat interval can be customized by @AngersZhuuuu in #170
- [ISSUE-171][BUG] LifeCycleManager throw cala.collection.immutable.HashMap$HashTrieMap cannot be cast to java.util.HashMap when handle destroyBuffersWithRetry by @AngersZhuuuu in #172
- [ISSUE-176][BUG] Handle RegisterWorker use wrong worker info when trigger lost event by @AngersZhuuuu in #177
- [ISSUE-178][BUG] RSS should not limit the io. numConnectionsPerPeer to 1, stability is not good enough by @AngersZhuuuu in #179
- [ISSUE-180][FEATURE] Disable rss.worker.closeIdleConnections by default by @AngersZhuuuu in #181
- [Feature] Multithread disk flusher by @FMX in #182
- Add instructions about restart cluster in HA mode. by @FMX in #189
- [DOC]Refine doc by @waitinfuture in #194
- [ISSUE-190][BUG] After WorkerLost, response to worker heartbeat RPC to, then worker can clean the data. by @AngersZhuuuu in #192
- [ISSUE-203][BUG] fix rss.master.address not work in rss-worker's configuration file issue by @dxheming in #204
- Cleanup unused code in network by @waitinfuture in #206
- [ISSUE-208] Refine log levels by @waitinfuture in #210
- [Feature] Support set rpc port config by @nafiyAix in #212
- [ISSUE-213] Cleanup unused code in RPC module by @waitinfuture in #214
- [ISSUE-217][FEATURE] Re-register worker after worker lost by @AngersZhuuuu in #219
Full Changelog: v0.1.0...v0.1.1