[
https://issues.apache.org/jira/browse/HBASE-7578?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13554784#comment-13554784
]
Lars Hofhansl commented on HBASE-7578:
--------------------------------------
This looks like the classical "missed signal" problem (although I find the code
in CatalogTracker a bit hard to track, so maybe I've missed it).
I just increase the wait-times in TestCatalogTracker a bit, this failure goes
away (or at least I did not see it after N*10 times as many runs, where N runs
would almost always produce this failure).
Increasing the timeouts is just a bandaid, though. I suppose we should follow
the classic pattern and have a "signalled" boolean that we set every time when
we notify, so that we do not lose the signaled state. Now if this was just a
test issue that would be overkill... But it looks like this a real issue where
the catalog tracker could indeed wait forever, because it missed the notifyAll.
> TestCatalogTracker hangs occasionally
> -------------------------------------
>
> Key: HBASE-7578
> URL: https://issues.apache.org/jira/browse/HBASE-7578
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Fix For: 0.94.5
>
> Attachments: 7578-jstack.txt
>
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
