[
https://issues.apache.org/jira/browse/PHOENIX-4990?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16661607#comment-16661607
]
Geoffrey Jacoby commented on PHOENIX-4990:
------------------------------------------
Maybe I'm missing something, but in the added test, shouldn't incrementAndGet
be called before the countDown on the latch, because the count down releases
the await() which will immediately check to see if the increment's happened?
Seems like a race condition.
> When stopping IndexWriter, give tasks a chance to complete
> ----------------------------------------------------------
>
> Key: PHOENIX-4990
> URL: https://issues.apache.org/jira/browse/PHOENIX-4990
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.14.0
> Reporter: Vincent Poon
> Assignee: Vincent Poon
> Priority: Major
> Attachments: PHOENIX-4990.v1.4.x-HBase-1.4.patch,
> PHOENIX-4990.v2.4.x-HBase-1.4.patch
>
>
> We've seen a race condition where an index write failure happens when a
> coprocessor is shutting down. Since we don't give the index writer threads a
> chance to complete when shutdownNow() is called, there is a chance the
> coprocessor shuts down its HTables while an index writer thread is in
> PhoenixIndexFailurePolicy trying to disable an index, which ends up using a
> closed HTable and received RejectedExecutionException
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
