[
https://issues.apache.org/jira/browse/MAPREDUCE-4157?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Lowe updated MAPREDUCE-4157:
----------------------------------
Attachment: MAPREDUCE-4157.patch
Patch to add a FINISHING application state as described for the initial patch
in MAPREDUCE-4099. Instead of transitioning from RUNNING to FINISHED, the
application transitions from RUNNING to FINISHING when it unregisters with the
RM. The RM then gives it so many seconds for the AM container to exit on its
own. If it fails to do so, it is killed. In either case, the app moves from
the FINISHING state to the FINISHED state.
The timeout for the FINISHING state is the same as the liveliness monitor
(i.e.: yarn.am.liveness-monitor.expiry-interval-ms) but we could easily add
another property if we think these need to be configured separately.
> ResourceManager should not kill apps that are well behaved
> ----------------------------------------------------------
>
> Key: MAPREDUCE-4157
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-4157
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: mrv2
> Affects Versions: 2.0.0
> Reporter: Jason Lowe
> Assignee: Jason Lowe
> Attachments: MAPREDUCE-4157.patch
>
>
> Currently when the ApplicationMaster unregisters with the ResourceManager,
> the RM kills (via the NMs) all the active containers for an application.
> This introduces a race where the AM may be trying to clean up and may not
> finish before it is killed. The RM should give the AM a chance to exit
> cleanly on its own rather than always race with a pending kill on shutdown.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
