[
https://issues.apache.org/jira/browse/ZOOKEEPER-1060?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13038657#comment-13038657
]
Jeremy Stribling commented on ZOOKEEPER-1060:
---------------------------------------------
I think this will also fix ZOOKEEPER-1069, which resulted in a corrupted
on-disk log. I am trying out the patch now, but we only saw the problem once in
months of testing, so it'll be hard for me to verify for sure.
> QuorumPeer takes a long time to shutdown
> ----------------------------------------
>
> Key: ZOOKEEPER-1060
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1060
> Project: ZooKeeper
> Issue Type: Bug
> Components: quorum
> Affects Versions: 3.4.0
> Reporter: Vishal K
> Assignee: Vishal K
> Priority: Minor
> Fix For: 3.4.0
>
> Attachments: ZOOKEEPER-1060.patch
>
>
> This problem is seen only if you have ZooKeeper embedded in your application.
> QuorumPeerMain.initializeAndRun() does a quorumPeer.join() before exiting.
> QuorumPeer.shutdown() tries to cleanup everything, but it does not interrupt
> itself. As a result, a if the peer is running FLE, it might be waiting to
> receive notifications (recvqueue.poll()) in FastLeaderElection. Therefore,
> quorumPeer.join() will wait until the peer wakes up from poll().
> The fix is simple - call this.interrupt() in QuorumPeer.shutdown().
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
