[
https://issues.apache.org/jira/browse/MAPREDUCE-697?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12912748#action_12912748
]
Greg Roelofs commented on MAPREDUCE-697:
----------------------------------------
I think the reporter is no longer around, but I'll ask anyway, just in case:
(1) Is the first issue still a problem? From inspection of the code, it
appears that it should work (that is, restarted trackers are considered
healthy, though their previous "faults" aren't erased).
(2) Clarification of the second issue: does "try to blacklist other TT" refer
to another TT on one of the previously blacklisted (25%) nodes or to a TT on
one of the never-blacklisted (75%) nodes?
> Jobwise list of blacklisted tasktrackers doesn't get refreshed even after
> restarting blacklisted tasktrackers.
> --------------------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-697
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-697
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: tasktracker
> Affects Versions: 0.20.1
> Reporter: Suman Sehgal
> Priority: Critical
>
> Jobwise list of blacklisted tasktrackers doesn't get refreshed even after
> restarting blacklisted tasktrackers. "jobdetails.jsp" page keeps on showing
> the same no. of blacklisted tasktrackers (it doesn't get back to zero).
> One associated issue:
> =================
> --> More than 25% of TTs are blacklisted in a job.
> --> Restart the blacklisted TTs. All the tasktrackers are healthy now.
> --> try to blacklist other TT for the same job.
> Not able to blacklist the "other" tasktracker even if
> "mapred.max.tracker.failures" exceeds the specified limit.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
