[
https://issues.apache.org/jira/browse/MAPREDUCE-5047?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13599295#comment-13599295
]
Sandy Ryza commented on MAPREDUCE-5047:
---------------------------------------
To elaborate on this a little further, IsolationRunner throws a NPE whenever it
starts due to not reading in mapred.local.dir from the job conf before trying
to use it. IsolationRunner has been broken for a long time and was removed in
MR2 (MAPREDUCE-2606).
While this patch doesn't fix IsolationRunner, keep.failed.task.files still has
utility without it, and this fixes the bug that was causing jobs to fail every
time it's turned on.
> keep.failed.task.files=true causes job failure on secure clusters
> -----------------------------------------------------------------
>
> Key: MAPREDUCE-5047
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5047
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: task, tasktracker
> Affects Versions: 1.1.1
> Reporter: Sandy Ryza
> Assignee: Sandy Ryza
> Fix For: 1.3.0
>
> Attachments: MAPREDUCE-5047.patch
>
>
> To support IsolationRunner, split info is written to local directories. This
> occurs inside MapTask#localizeConfiguration, which is called both tasktracker
> and by the child JVM. On a secure cluster, the tasktacker's attempt to write
> it fails, because the tasktracker does not have permission to write to the
> user's directory. It is likely that the call to localizeConfiguration in the
> tasktracker can be removed.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
