[
https://issues.apache.org/jira/browse/OOZIE-1767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13960485#comment-13960485
]
Mona Chitnis commented on OOZIE-1767:
-------------------------------------
The idea behind not having this functionality was to discourage users from
using java action to spawn own MR jobs and use the map-reduce action instead. I
see that in your use case of using other frameworks like Crunch/Cascading you'd
need to use some custom logic before launching jobs - so map-reduce action
won't be enough.
We don't scrape logs for all action types. For pig action and map-reduce
action, we use the PigStats and JobClient API respectively to get the actual
job ids. Logic used by hive-action might actually be sub-optimal.
> Java Action automatically record child job ids
> ----------------------------------------------
>
> Key: OOZIE-1767
> URL: https://issues.apache.org/jira/browse/OOZIE-1767
> Project: Oozie
> Issue Type: Improvement
> Components: action
> Reporter: Micah Whitacre
>
> It is common for consumers to utilize Big Data frameworks like Crunch and
> Cascading inside of their larger processing workflows. These frameworks when
> ran will launch potentially one to many MR jobs. The common way to launch
> these is using the Java action as both frameworks have a specific API for
> launching and executing the jobs. Unfortunately for the default Java action,
> those spawned processes are not tied to the Oozie action which launched them.
> It would be great if there was a way for jobs launched from the Java action
> to be tied back to the Oozie action.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
