[
https://issues.apache.org/jira/browse/MAPREDUCE-5875?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13991020#comment-13991020
]
Gera Shegalov commented on MAPREDUCE-5875:
------------------------------------------
v02 in this JIRA does not assume mapred-site.xml on the class path. It covers
this general scenario (modulo failing tests :) ).
# client 1 submits a job with conf 1. Its customizations + cluster
customizations result in job.xml on the cluster.
# client 2 connects to the cluster and calls getJob. Now even if it had slight
different conf2, the job object returned will be the one from step 1. I don't
think that MAPREDUCE-5856 correctly deals with this
> Make Counter limits consistent conf across JobClient, MRAppMaster, and
> YarnChild
> --------------------------------------------------------------------------------
>
> Key: MAPREDUCE-5875
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5875
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: applicationmaster, client, task
> Affects Versions: 2.4.0
> Reporter: Gera Shegalov
> Assignee: Gera Shegalov
> Attachments: MAPREDUCE-5875.v01.patch, MAPREDUCE-5875.v02.patch
>
>
> Currently, counter limits "mapreduce.job.counters.*" handled by
> {{org.apache.hadoop.mapreduce.counters.Limits}} are initialized
> asymmetrically: on the client side, and on the AM, job.xml is ignored whereas
> it's taken into account in YarnChild.
> It would be good to make the Limits job-configurable, such that max
> counters/groups is only increased when needed. With the current Limits
> implementation relying on static constants, it's going to be challenging for
> tools that submit jobs concurrently without resorting to class loading
> isolation.
> The patch that I am uploading is not perfect but demonstrates the issue.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
