[
https://issues.apache.org/jira/browse/HADOOP-1226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12498078
]
Arun C Murthy commented on HADOOP-1226:
---------------------------------------
<code>
private Path getTaskOutputPath(JobConf conf) {
return new Path(conf.getOutputPath(), ("_" + taskId));
}
</code>
That is how taskOutputPath is created, and thus should be an absolute path. I'm
not quite sure, looking at TestMiniMRWithDFS.launchWordCount atleast, how this
could result in a relative path... not-fully-qualified maybe, but not relative.
Anyway, the short answer is making taskOutputPath fully-qualified shouldn't be
a problem.
> Also, taskOutputPath stuff is reduce specific, so if anywhere, it belongs in
> ReduceTask, no?
This also comes into play with 'reducer -NONE' where map outputs need to go
directly to dfs...
Urelated note: +1 to moving that code from mapred kernel to the
OutputFormatBase.
> makeQualified should return an instance of a DfsPath when passed a DfsPath
> ----------------------------------------------------------------------------
>
> Key: HADOOP-1226
> URL: https://issues.apache.org/jira/browse/HADOOP-1226
> Project: Hadoop
> Issue Type: Bug
> Components: dfs
> Reporter: Koji Noguchi
> Assigned To: dhruba borthakur
> Fix For: 0.14.0
>
> Attachments: makeQualified2.patch
>
>
> Currently, it returns a instance of a Path.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
