[
https://issues.apache.org/jira/browse/HBASE-14378?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14734964#comment-14734964
]
stack commented on HBASE-14378:
-------------------------------
bq. I thought we were timing out based on test categorization now? Or do these
need more time than the default?
The timeout by category requires instrumenting each test class HBASE-14356 (You
have to add a junit Rule -- see parent of HBASE-1435 patch for example).
At head of this issue I remark that timeouts do not apply to teardown phase --
bummer.
I'd started in on adding timeouts so just ran it out to the end.
Thanks for taking a look at these zombies Sean. Let me try adding in same
restore of priority handlers to see if it helps on these zombies.
> Get TestAccessController* passing again on branch-1
> ---------------------------------------------------
>
> Key: HBASE-14378
> URL: https://issues.apache.org/jira/browse/HBASE-14378
> Project: HBase
> Issue Type: Sub-task
> Reporter: stack
> Assignee: stack
> Fix For: 2.0.0, 1.2.0, 1.3.0
>
> Attachments: 14378.branch-1.txt
>
>
> TestAccessController* are failing reliably on branch-1. They go zombie. I
> learned that setting the junit test timeout facility on the class doesn't
> make the zombie timeout nor does setting a timeout on each test turn zombies
> to test failures; the test goes zombie on the way out in the tear down of the
> cluster.
> Digging, we are out of handlers... all are occupied.
> 3dacee6 HBASE-14290 Spin up less threads in tests cut the default thread
> count to 3 from 10. Putting the value back on these tests seems to make them
> pass reliably when I run locally. For good measure, I'll add in the timeouts
> .
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
