[
https://issues.apache.org/jira/browse/MAPREDUCE-779?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740387#action_12740387
]
Hemanth Yamijala commented on MAPREDUCE-779:
--------------------------------------------
I looked at the patch. Overall, looking good.
Sharad and I feel that the method
JobTrackerStatistics.taskTrackerHealthCheckFailed() is not required. I think it
is assumed that the tasktracker stat will be created and available. Further, it
seems like the class JobTrackerStatistics manages the TasktrackerStatistics
instances and nothing more. So, the method seems out of place. Hence, wherever
we need to update the health check failure count, we could simply call
JobTrackerStatistics.getTaskTrackerStat().incrHealthCheckFailed(). Sreekanth,
any reason you thought the TaskTrackerStat for a TT will not be available and
hence were implicitly managing the creation of it ?
One more minor nit is we can rename the label "Blacklisted due to health
Failures" to "Failed Health Checks" just to reduce the verbosity.
Other than this +1
> Add node health failures into JobTrackerStatistics
> --------------------------------------------------
>
> Key: MAPREDUCE-779
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-779
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: jobtracker
> Reporter: Sreekanth Ramakrishnan
> Assignee: Sreekanth Ramakrishnan
> Attachments: mapreduce-779-1.patch, mapreduce-779-2.patch,
> mapreduce-779-3.patch
>
>
> Add the node health failure counts into {{JobTrackerStatistics}}.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
