[
https://issues.apache.org/jira/browse/MAPREDUCE-5773?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Lowe updated MAPREDUCE-5773:
----------------------------------
Resolution: Fixed
Fix Version/s: 2.4.0
Hadoop Flags: Reviewed
Status: Resolved (was: Patch Available)
Thanks, Gera! I committed this to trunk, branch-2, and branch-2.4.
> Provide dedicated MRAppMaster syslog length limit
> -------------------------------------------------
>
> Key: MAPREDUCE-5773
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5773
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: mr-am
> Affects Versions: 2.3.0
> Reporter: Gera Shegalov
> Assignee: Gera Shegalov
> Priority: Blocker
> Fix For: 2.4.0
>
> Attachments: MAPREDUCE-5773.v01.patch
>
>
> Currently mapreduce.task.userlog.limit.kb controls both the length of task
> attempt logs and MA-AM attempt logs. Obviously, MR-AM log is not userlog.
> We are interested in keeping MR-AM log either in its entirety or in much
> larger size than task logs for debugging.
> MAPREDUCE-5672 introduced CRLA. We already use a dedicated setting for how
> many backups of rolled files we keep for MR-AM. However, for large jobs with
> tens of megabyte AM log, it means that you have a long series of files. A
> natural improvement is to have a configuration
> yarn.app.mapreduce.am.container.log.limit.kb . It allows to either disable
> rolling the log for altogether while keeping it for the task attempts, or to
> use much a larger limit to make rolling less frequent.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
