[
https://issues.apache.org/jira/browse/MAPREDUCE-6856?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15901505#comment-15901505
]
Gergő Pásztor commented on MAPREDUCE-6856:
------------------------------------------
This problem was solved with the MAPREDUCE-6839's patch.
> TestRecovery.testSpeculative fails if testCrashed fails
> -------------------------------------------------------
>
> Key: MAPREDUCE-6856
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6856
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Peter Bacsko
> Assignee: Peter Bacsko
>
> The test {{testSpeculative}} in
> {{org.apache.hadoop.mapreduce.v2.app.TestRecovery}} is unstable.
> Based on my findings, the test itself is not problematic. It only fails if
> {{testCrashed}} in the same class fails before it.
> The reason is not completely clear to me, but I whenever I explicitly stop
> the MRAppMaster in {{testCrashed}} in a finally block, then the issue
> disappears. I think the reason is that both tests uses the same folder for
> staging.
> Solution: wrap logic in {{testCrashed}} in a try-finally block and then stop
> the MRAppMaster.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
