[ https://issues.apache.org/jira/browse/HBASE-19290?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16263720#comment-16263720 ]
binlijin commented on HBASE-19290: ---------------------------------- [~tedyu] bq. Assuming patch v3 is very close to the version you run in the 2000+ node production cluster, can you post some performance numbers (in terms of reduction in zookeeper requests) so that we can know its effectiveness ? We do not record the performance numbers. But without the patch we can see the HMaster get the zookeeper event very very slowly... HMaster put up split task: *2017-07-11 20:22:57,608* DEBUG [main-EventThread] coordination.SplitLogManagerCoordination: put up splitlog task at znode /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 RegionServer grab the task and done it. *2017-07-11 20:23:33,689* INFO [SplitLogWorker-hadoop1435:16020] coordination.ZkSplitLogWorkerCoordination: worker hadoop1435.et2.tbsite.net,16020,1495647366458 acquired task /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 *2017-07-11 20:25:47,131* INFO [RS_LOG_REPLAY_OPS-hadoop1435:16020-1] coordination.ZkSplitLogWorkerCoordination: successfully transitioned task /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 to final state DONE hadoop1435.et2.tbsite.net,16020,1495647366458 HMaster get the task done event and delete it: *2017-07-11 20:49:52,879* INFO [main-EventThread] coordination.SplitLogManagerCoordination: task /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 entered state: DONE hadoop1435.et2.tbsite.net,16020,1495647366458 *2017-07-11 20:49:52,881* INFO [main-EventThread] coordination.SplitLogManagerCoordination: Done splitting /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 *2017-07-11 21:19:52,280* DEBUG [main-EventThread] coordination.ZKSplitLogManagerCoordination$DeleteAsyncCallback: deleted /hbase/splitWAL/WALs%2Fhadoop0448.et2.tbsite.net%2C16020%2C1495647366007-splitting%2Fhadoop0448.et2.tbsite.net%252C16020%252C1495647366007.regiongroup-2.1499768090548 > Reduce zk request when doing split log > -------------------------------------- > > Key: HBASE-19290 > URL: https://issues.apache.org/jira/browse/HBASE-19290 > Project: HBase > Issue Type: Improvement > Reporter: binlijin > Assignee: binlijin > Attachments: HBASE-19290.master.001.patch, > HBASE-19290.master.002.patch, HBASE-19290.master.003.patch, > HBASE-19290.master.004.patch > > > We observe once the cluster has 1000+ nodes and when hundreds of nodes abort > and doing split log, the split is very very slow, and we find the > regionserver and master wait on the zookeeper response, so we need to reduce > zookeeper request and pressure for big cluster. > (1) Reduce request to rsZNode, every time calculateAvailableSplitters will > get rsZNode's children from zookeeper, when cluster is huge, this is heavy. > This patch reduce the request. > (2) When the regionserver has max split tasks running, it may still trying to > grab task and issue zookeeper request, we should sleep and wait until we can > grab tasks again. -- This message was sent by Atlassian JIRA (v6.4.14#64029)