Release 0.9.5.1

1. Add storm.messaging.netty.async.block setting

0.9.5.1-rc

1. Add netty sync mode
2. Add block operation in netty async mode
3. Replace exception with Throwable in executor layer
4. Upgrade curator-framework version from 1.15 to 1.3.2
5. Add more netty junit test
6. Add log when queue is full

0.9.5 new JStorm Scheduler

Big feature:

1. Redesign scheduler arithmetic, basing storm scheduler, but more smarter, more tuples will be sent in one worker.

Bug fix

1. Fix disruptor use too much cpu
2. Add target NettyServer log when fail to send data by netty

0.9.4.1 release

Bug fix:

1. Improve speed between tasks who is running in one worker
2. Fix wrong timeout seconds
3. Add checking port when worker initialize and begin to kill old worker
4. Move worker hearbeat thread before initializing tasks
5. Move init netty-server before initializeing tasks
6. Check whether tuple's rootId is duplicated
7. Add default value into Utils.getInt
8. Add result function in ReconnectRunnable
9. Add operation to start Timetick
10. Halt process when master nimbus lost ZK node
11. Add exception catch when cgroups kill process
12. Speed up reconnect to netty-server
13. Share one task hearbeat thread for all tasks
14. Quickly haltprocess when initialization failed.
15. Check web-ui logview page size

0.9.4 release

Big features

1. Add transaction programming mode
2. Rewrite netty code
   • use share boss/worker thread pool
   • async send batch tuples
   • single thread to do reconnect job
   • receive batch tuples
3. Add metrics and statics
4. Merge Alimama storm branch into this version, submit jar with -conf, -D, -lib

Enhancement

1. add setting when supervisor has been shutdown, worker will shutdown automatically
2. add LocalFristGrouping api
3. enable cgroup for normal user

Bug fix:

1. Setting buffer size when upload jar
2. Add lock between ZK watch and timer thread when refresh connection
3. Enable nimbus monitor thread only when topology is running in cluster mode
4. Fix exception when failed to read old assignment of ZK
5. classloader fix when both parent and current classloader load the same class
6. Fix log view null pointer exception

0.9.3.1 release

Enhancement

1. Switch apache thrift7 to storm thrift7
2. Set defatult acker number is 1
3. Add "spout.single.thread" setting
4. Make nimbus logview port different from supervisor's
5. Web ui can list all files of log's subdir
6. Set gc dump dir as log's dir

0.9.3.1 RC

Big Change

1. Switch apache thrift7 to storm thrift7

Enhancement

1. Set defatult acker number is 1
2. Add "spout.single.thread" setting

0.9.3 Release

New feature

1. Support Aliyun Apsara/Hadoop Yarn

Enhancement

1. Redesign Logview
2. Kill old worker under the same port when worker is starting
3. Add zk information/version information on UI
4. Add nodeport information for dead task in nimbus
5. Add interface to get values when spout doing ack
6. Add timeout statics in bolt
7. jstorm script return status
8. Add logs when fail to deserialize tuple
9. Skip sleep operation when max_pending is 1 and waiting ack
10. Remove useless dependency
11. Longer task timeout setting
12. Add supervisor.use.ip setting
13. Redirect supervisor out/err to /dev/null, redirect worker out/err to one file

Bug Fix

1. Fix kryo fail to deserialize object when enable classloader
2. Fix fail to reassign dead task when worker number is less than topology apply
3. Set samller jvm heap memory for jstorm-client
4. Fix fail to set topology status as active when do rebalance operation twice at one time,
5. Fix local mode bug under linux
6. Fix average latency isn't accurate
7. GC tuning.
8. Add default kill function for AysncLoopRunnable

0.9.2 release

New feature

1. Support LocalCluster/LocalDrpc mode, support debugging topology under local mode
2. Support CGroups, assigning CPU in hardware level.
3. Support simple logview

Bug fix

1. Change SpoutExecutor's RotatingMap to TimeCacheMap, when putting too much timeout tuple is easy to cause deadlock in spout acker thread
2. Tunning gc parameter, improve performance and avoid full GC
3. Improve Topology's own gc priority, make it higher than JStorm system setting.
4. Tuning Nimbus HA, switch nimbus faster, when occur nimbus failure.
5. Fix bugs found by FindBugs tool.
6. Revert Trident interface to 0.8.1, due to 0.8.1's trident interface's performance is better.
7. Setting nimbus.task.timeout.secs as 60 to avoid nimbus doing assignment when task is under full gc.
8. Setting default rpc framework as netty
9. Tunning nimbus shutdown flow
10. Tunning worker shutdown flow
11. Add task heartbeat log
12. Optimize Drpc/LocalDrpc source code.
13. Move classloader to client jar.
    14 Fix classloader fail to load anonymous class
14. Web Ui display slave nimbus
15. Add thrift max read buffer size
16. Setting CPU slot base double
17. Move Zk utility to jstorm-client-extension.jar
18. Fix localOrShuffle null pointer
19. Redirecting worker's System.out/System.err to file is configurable.
20. Add new RPC frameworker JeroMq
21. Fix Zk watcher miss problem
22. Update sl4j 1.5.6 to 1.7.5
23. Shutdown worker when occur exception in Smart thread
24. Skip downloading useless topology in Supervisor
25. Redownload the topology when failed to deserialize topology in Supervisor.
26. Fix topology codeDir as resourceDir
27. Catch error when normalize topology
28. Add log when found one task is dead
29. Add maven repository, JStorm is able to build outside of Alibaba
30. Fix localOrShuffle null pointer exception
31. Add statics counting for internal tuples in one worker
32. Add thrift.close after download topology binary in Supervisor

0.9.1

New feature

1. Application classloader. when Application jar is conflict with jstorm jar,
   please enable application classloader.
2. Group Quato, Different group with different resource quato.

Bug fix

1. Fix Rotation Map competition issue.
2. Set default acker number as 0
3. Set default spout/bolt number as 1
4. Add log directory in log4j configuration file
5. Add transaction example
6. Fix UI showing wrong worker numbe in topology page
7. Fix UI showing wrong latency in topology page
8. Replace hardcode Integer convert with JStormUtils.parseInt
9. Support string parse in Utils.getInt
10. Remove useless dependency in pom.xml
11. Support supervisor using IP or special hostname
12. Add more details when no resource has been assigned to one new topology
13. Replace normal thread with Smart thread
14. Add gc details
15. Code format
16. Unify stormId and topologyId as topologyId
17. Every nimbus will regist ip to ZK