[
https://issues.apache.org/jira/browse/HBASE-9948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13865848#comment-13865848
]
Gustavo Anatoly commented on HBASE-9948:
----------------------------------------
Hi, [~jeffreyz].
I will follow your suggestions and really to avoid data loss the request
splitting log process should be an atomic operation, so the best way is
investigate the root causes of dup log.
[~yuzhih...@gmail.com], How can I reproduce this scenario?
Thank you [~jeffreyz].
> HMaster should handle duplicate log split requests
> --------------------------------------------------
>
> Key: HBASE-9948
> URL: https://issues.apache.org/jira/browse/HBASE-9948
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Gustavo Anatoly
> Attachments: HBASE-9948-v2.patch, HBASE-9948.patch
>
>
> I saw the following in test output for TestRestartCluster:
> {code}
> 2013-11-11 19:59:55,538 DEBUG [M:0;kiyo:36213] master.SplitLogManager(327):
> Scheduling batch of logs to split
> 2013-11-11 19:59:55,538 INFO [M:0;kiyo:36213] master.SplitLogManager(329):
> started splitting 1 logs in
> [hdfs://localhost:46376/user/hortonzy/hbase/WALs/kk,44962,1384199990193-splitting]
> 2013-11-11 19:59:55,538 WARN [M:0;kiyo:36213] master.SplitLogManager(1048):
> Failure because two threads can't wait for the same task;
> path=/hbase/splitWAL/WALs%2Fkk%2C44962%2C1384199990193-splitting%2Fkk%252C44962%252C1384199990193.1384199993702.meta
> 2013-11-11 19:59:55,538 FATAL [M:0;kiyo:36213] master.HMaster(2188): Master
> server abort: loaded coprocessors are:
> [org.apache.hadoop.hbase.coprocessor.MultiRowMutationEndpoint]
> 2013-11-11 19:59:55,538 FATAL [M:0;kiyo:36213] master.HMaster(2193):
> Unhandled exception. Starting shutdown.
> java.io.IOException: duplicate log split scheduled for
> hdfs://localhost:46376/user/hortonzy/hbase/WALs/kk,44962,1384199990193-splitting/kk%2C44962%2C1384199990193.1384199993702.meta
> at
> org.apache.hadoop.hbase.master.SplitLogManager.splitLogDistributed(SplitLogManager.java:343)
> at
> org.apache.hadoop.hbase.master.MasterFileSystem.splitLog(MasterFileSystem.java:409)
> at
> org.apache.hadoop.hbase.master.MasterFileSystem.splitMetaLog(MasterFileSystem.java:301)
> at
> org.apache.hadoop.hbase.master.MasterFileSystem.splitMetaLog(MasterFileSystem.java:292)
> at
> org.apache.hadoop.hbase.master.HMaster.assignMeta(HMaster.java:1038)
> at
> org.apache.hadoop.hbase.master.HMaster.finishInitialization(HMaster.java:868)
> at org.apache.hadoop.hbase.master.HMaster.run(HMaster.java:605)
> at java.lang.Thread.run(Thread.java:724)
> 2013-11-11 19:59:55,539 INFO [M:0;kiyo:36213] master.HMaster(2386): Aborting
> 2013-11-11 19:59:55,539 DEBUG [M:0;kiyo:36213] master.HMaster(1234): Stopping
> service threads
> {code}
> HMaster should handle duplicate log split requests, instead of aborting.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
