[
https://issues.apache.org/jira/browse/MAPREDUCE-5483?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13751574#comment-13751574
]
Chuan Liu commented on MAPREDUCE-5483:
--------------------------------------
I want to understand the scenario a little bit more. After reverting this
change, it means:
# If the directory already exists, we will enforce it belongs to the current
user.
# Otherwise, we create a new directory. Ownership does not matter in this case.
> revert MAPREDUCE-5357
> ---------------------
>
> Key: MAPREDUCE-5483
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5483
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: distcp
> Affects Versions: 2.1.0-beta
> Reporter: Alejandro Abdelnur
> Assignee: Robert Kanter
> Fix For: 2.1.1-beta
>
> Attachments: MAPREDUCE-5483.patch
>
>
> MAPREDUCE-5357 does a fileystem chown() operation. chown() is not valid
> unless you are superuser. if you a chown() to yourself is a NOP, that is why
> has not been detected in Hadoop testcases where user is running as itself.
> However, in distcp testcases run by Oozie which use test users/groups from
> UGI for minicluster it is failing because of this chown() either because the
> test user does not exist of because the current use does not have privileges
> to do a chown().
> We should revert MAPREDUCE-5357. Windows should handle this with some
> conditional logic used only when running in Windows.
> Opening a new JIRA and not reverting directly because MAPREDUCE-5357 went in
> 2.1.0-beta.
--
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