You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Gera Shegalov (JIRA)" <ji...@apache.org> on 2014/09/15 05:42:33 UTC

[jira] [Commented] (MAPREDUCE-5929) YARNRunner.java, path for jobJarPath not set correctly

    [ https://issues.apache.org/jira/browse/MAPREDUCE-5929?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14133543#comment-14133543 ] 

Gera Shegalov commented on MAPREDUCE-5929:
------------------------------------------

Hi [~rpalamut], thanks for the patch. Can you add a test to TestMRJobs where staging directory is on a local file system or on a different mini dfs cluster?

> YARNRunner.java, path for jobJarPath not set correctly
> ------------------------------------------------------
>
>                 Key: MAPREDUCE-5929
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5929
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>    Affects Versions: 2.2.0
>            Reporter: Chao Tian
>            Assignee: Rahul Palamuttam
>              Labels: newbie, patch
>         Attachments: MAPREDUCE-5929.patch
>
>
> In YARNRunner.java, line 357,
> Path jobJarPath = new Path(jobConf.get(MRJobConfig.JAR));
> This causes the job.jar file to miss scheme, host and port number on distributed file systems other than hdfs. 
> If we compare line 357 with line 344, there "job.xml" is actually set as
>  
> Path jobConfPath = new Path(jobSubmitDir,MRJobConfig.JOB_CONF_FILE);
> It appears "jobSubmitDir" is missing on line 357, which causes this problem. In hdfs, the additional qualify process will correct this problem, but not other generic distributed file systems.
> The proposed change is to replace 35 7 with
> Path jobJarPath = new Path(jobConf.get(jobSubmitDir,MRJobConfig.JAR));
>  



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)