[
https://issues.apache.org/jira/browse/MAPREDUCE-2207?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12975398#action_12975398
]
Scott Chen commented on MAPREDUCE-2207:
---------------------------------------
Hi Liyin,
Thanks for the work. Taking a second look at this one, I think it may be better
to move this logic to server side.
We can call hasFailedAndNeedCleanupTaskToReport() in
JobTracker.getSetupAndCleanupTasks().
This way the TT can still get other tasks from taskScheduler.assignTasks() and
immediately run tasks without waiting the next heartbeat.
What do you think?
> Task-cleanup task should not be scheduled on the node that the task just
> failed
> -------------------------------------------------------------------------------
>
> Key: MAPREDUCE-2207
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2207
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: jobtracker
> Affects Versions: 0.23.0
> Reporter: Scott Chen
> Fix For: 0.23.0
>
> Attachments: 0.19.1.diff, 2207-1.diff
>
>
> Currently the task-cleanup task always go to the same node that the task just
> failed.
> There is a higher chance that it hits a bad node. This should be changed.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
