[jira] [Commented] (MAPREDUCE-2350) LocalJobRunner uses "mapred.output.committer.class" configuration property to retrive the OutputCommitter (regardless of whether the old API is used or the new API)

["Shai Erera (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MAPREDUCE-2350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13078844#comment-13078844 ] 

Shai Erera commented on MAPREDUCE-2350:
---------------------------------------

Thanks Devaraj for looking into this. Patch looks good to me.

> LocalJobRunner uses "mapred.output.committer.class" configuration property to retrive the OutputCommitter (regardless of whether the old API is used or the new API)
> --------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-2350
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-2350
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Ahmed Radwan
>            Assignee: Devaraj K
>             Fix For: 0.23.0
>
>         Attachments: MAPREDUCE-2350.patch
>
>
> LocalJobRunner uses the "mapred.output.committer.class" configuration property to retrieve the output committer for the job, which can be different from the Output Committer returned from OutputFormat.getOutputCommitter(TaskAttemptContext context). So, two different output committers can be used in the same job.
> See line 324 in org.apache.hadoop.mapred.LocalJobRunner: OutputCommitter outputCommitter = job.getOutputCommitter();
> Need to modify this behavior to check if the new or the old API is used, and then return the correct output committer. 

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira