You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Mikhail Davidov (JIRA)" <ji...@apache.org> on 2013/10/14 02:03:41 UTC
[jira] [Commented] (MAPREDUCE-5582) Setting
mapred.job.reduce.memory.mb to 0 from a job with CapacityTracker leads to
inconsistent state in JVMManager
[ https://issues.apache.org/jira/browse/MAPREDUCE-5582?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13793851#comment-13793851 ]
Mikhail Davidov commented on MAPREDUCE-5582:
--------------------------------------------
I believe this also affects mapred.job.map.memory.mb as well but have not tested.
> Setting mapred.job.reduce.memory.mb to 0 from a job with CapacityTracker leads to inconsistent state in JVMManager
> ------------------------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-5582
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5582
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: capacity-sched
> Reporter: Mikhail Davidov
> Priority: Critical
>
> If a job sets mapred.job.reduce.memory.mb to 0 the capacity scheduler incorrectly allocates resources eventually causing "Inconsistent state!!! JVM Manager reached an unstable state while reaping a JVM for task" errors from the JVMManager killing all TaskTrackers that have been used for the job.
--
This message was sent by Atlassian JIRA
(v6.1#6144)