[
https://issues.apache.org/jira/browse/HADOOP-4287?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12637421#action_12637421
]
Sreekanth Ramakrishnan commented on HADOOP-4287:
------------------------------------------------
The counter assignments from terminate method should remain, because only thing
which we are handling is prevention of updating the counters after the job has
been terminated due to maximum failures reached, as mentioned by Vinod above.
If we remove the assignment then value of running reduces would not fall back
to zero, in case a job which is terminated when it has 5 maps in total and has
2 running maps when it is terminated. Then if we remove the assignment from
terminate then the JIP.running task would say 2 and waiting would always point
to 3.
> [mapred] jobqueue_details.jsp shows negative count of running and waiting
> reduces with CapacityTaskScheduler.
> -------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-4287
> URL: https://issues.apache.org/jira/browse/HADOOP-4287
> Project: Hadoop Core
> Issue Type: Bug
> Components: contrib/capacity-sched
> Reporter: Vinod K V
> Assignee: Sreekanth Ramakrishnan
> Priority: Blocker
> Attachments: HADOOP-4287-1.patch
>
>
> This I observed while running a job that always fails because of reduce
> failures. Need to investigate this.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
