[
https://issues.apache.org/jira/browse/MAPREDUCE-1659?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12859148#action_12859148
]
Rodrigo Schmidt commented on MAPREDUCE-1659:
--------------------------------------------
hadoop.tmp.dir seems to be the local tmp directory used by mapreduce. Maybe
it's better to leave "/tmp" in that case. Anyway, there are config vars for
these temp dirs at the raid level, so users can redirect the tmp files to
different places if they wish to.
> RaidNode should write temp files on /tmp and add random numbers to their
> names to avoid conflicts
> -------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-1659
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-1659
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: contrib/raid
> Affects Versions: 0.22.0
> Reporter: Rodrigo Schmidt
> Assignee: Rodrigo Schmidt
> Fix For: 0.22.0
>
> Attachments: MAPREDUCE-1659.0.patch
>
>
> The RaidNode methods to raid files and recover them should write recovery and
> tmp files on /tmp instead of /raid.
> Besides that, filenames should have a random number appended to them to avoid
> conflicts. This makes the code safer and avoids errors when multiple
> recoveries run in parallel.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
