[
https://issues.apache.org/jira/browse/OOZIE-1452?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13709208#comment-13709208
]
Virag Kothari commented on OOZIE-1452:
--------------------------------------
After discussion with Rohini and Mohammad, keeping the JIRA open and cancelling
the patch bcoz:
1) If CoordActionUpdateX acquires action lock instead of job lock, then it
might create inconsistency as coordkill, coordsuspend etc. acquire job lock and
can then execute parallel with coordactionupdateX
2) OTOH, if we do the reverse and make CoordActionCheckx acquire job lock
instead of action lock, the system will be in consistent state as all coord
related commands acquire job lock now. But it makes the overall system slower
as some of the coord commands which only update coord action have still to
acquire a more granular lock(job) and creates lock contention. This is
especially seen in stress conditions with HCAT testing.
> CoordActionUpdateX should acquire lock for coord action and not coord job
> -------------------------------------------------------------------------
>
> Key: OOZIE-1452
> URL: https://issues.apache.org/jira/browse/OOZIE-1452
> Project: Oozie
> Issue Type: Bug
> Reporter: Virag Kothari
> Assignee: Virag Kothari
> Attachments: OOZIE-1452.patch
>
>
> CoordActionCheck acquires lock for action while Coordactionupdate acquire
> lock for coord job. Hence they were seen executing at the same time. This can
> cause problems like unnecessary database calls and duplicate notifications.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira