[
https://issues.apache.org/jira/browse/OOZIE-1964?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robert Kanter updated OOZIE-1964:
---------------------------------
Attachment: OOZIE-1964.patch
BeeLine prints the Hadoop Job ID, among other MR information, to stderr. The
patch duplicates the error stream that BeeLine uses to a file, so that we can
harvest the Job ID, just like we do with most of the other actions. Due to how
that works, I had to actually create a {{BeeLine}} object and call a couple
methods instead of simply calling {{BeeLine.main}}.
The patch also refactors some common code for writing the Hadoop Job IDs
between {{SqoopMain}}, {{HiveMain}}, {{PigMain}}, and now {{Hive2Main}} into
{{LauncherMain}}. It also updates the Oozie web console to be able to display
the child IDs for the Hive2 action (and also the Sqoop action, which was
missing).
Versions of Hive with HIVE-7615 will make BeeLine print the MR information,
including the Hadoop Job ID, to stderr. In older versions, they won't be
harvested, but nothing bad will happen.
I tested that all of the above actions still work correctly in a cluster. And
I verified that the Hadoop Job IDs are harvested from the Hive2 action by using
a newer version of Hive that has HIVE-7615.
> Hive Server 2 action doesn't return Hadoop Job IDs
> --------------------------------------------------
>
> Key: OOZIE-1964
> URL: https://issues.apache.org/jira/browse/OOZIE-1964
> Project: Oozie
> Issue Type: Sub-task
> Components: action
> Affects Versions: trunk
> Reporter: Robert Kanter
> Assignee: Robert Kanter
> Fix For: trunk
>
> Attachments: OOZIE-1964.patch
>
>
> Beeline currently doesn't support getting the Hadoop Job IDs for jobs
> launched by Hive Server 2. When/If Beeline ever adds support for this, we
> should update the Hive Server 2 action to parse the IDs and return them back
> to the Oozie server like most actions do.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
