[
https://issues.apache.org/jira/browse/MAPREDUCE-5154?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13646202#comment-13646202
]
Sandy Ryza commented on MAPREDUCE-5154:
---------------------------------------
I considered that, and my thought for the separate async call was that I didn't
see the token cancellation as related to deletions in a general sense. But
your approach seems reasonable to me too, and has the advantage of stirring up
less code, so +1.
Nit: it might be good to add a comment to the new PathDeletionContext that
explains what it's used for and makes it clear that the token cancellation
comes after the deletion.
> staging directory deletion fails because delegation tokens have been cancelled
> ------------------------------------------------------------------------------
>
> Key: MAPREDUCE-5154
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5154
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: jobtracker
> Affects Versions: 1.2.0
> Reporter: Sandy Ryza
> Assignee: Sandy Ryza
> Attachments: MAPREDUCE-5154.patch, MAPREDUCE-5154.patch
>
>
> In a secure setup, the jobtracker needs the job's delegation tokens to delete
> the staging directory. MAPREDUCE-4850 made it so that job cleanup staging
> directory deletion occurs asynchronously, so that it could order it with
> system directory deletion. This introduced the issue that a job's delegation
> tokens could be cancelled before the cleanup thread got around to deleting
> it, causing the deletion to fail.
--
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
