[
https://issues.apache.org/jira/browse/OOZIE-381?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13101822#comment-13101822
]
Hadoop QA commented on OOZIE-381:
---------------------------------
deepikakhera remarked:
Hi Mohammad,
I needed the feature now, so I did a quick patch to get it working (did it for
version 2.2 for oozie. Now , I plan to use it for 2.3). Basically the issue
was that JavaActionExecutor.getActionFileSystem(Context context, Element
actionXml), is not using the filesystem from actionXml (which has the
configuration for the second job that is to be run in the other
cluster).Instead it uses the filesystem of workflow.
The exception I got (when tested with version 2.2) is:
IllegalArgumentException: Wrong FS:
hdfs://cluster1namenode:9000/user/Deepika/oozie-root/0000006-110201204059448-oozie-root-W/cluster1job1--map-reduce,
expected: hdfs://cluster2namenode:9000.
The workflow directory here is cluster1. The exception is thrown when the
workflow tries to run a map reduce job in cluster2.
In my patch, I modified the getActionFileSystem method to use the filesystem
from the actionXml for "map reduce jobs", when that is present.
I can submit the patch, if you think it will be useful - until the permanent
fix is out in version 3.0.
Thanks,
Deepika
> GH-616: Unable to run a map reduce job in a cluster different from where
> Oozie is running.
> ------------------------------------------------------------------------------------------
>
> Key: OOZIE-381
> URL: https://issues.apache.org/jira/browse/OOZIE-381
> Project: Oozie
> Issue Type: Bug
> Reporter: Hadoop QA
>
> If a map reduce job is configured to run on a cluster different from the
> cluster on which Oozie is running, it fails with error.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
