[
https://issues.apache.org/jira/browse/YARN-5292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15731526#comment-15731526
]
Hitesh Sharma commented on YARN-5292:
-------------------------------------
Hi [~kasha], the design doc and the patch allow an opportunistic container to
be paused and resumed. The actual implementation of the pluggable interface
will be as part of [YARN-5196] and there via yarn-site.xml you can specify the
preemption policy for an opportunistic container when there is a guaranteed
container waiting to run (i.e. pause or kill). Currently only the NM scheduler
can raise the PAUSE/RESUME events and there is no support for doing so in the
container management protocol.
One of the questions on the table is if there is a need to extend support of
PAUSE/RESUME to guaranteed containers and have the AMs initiate that. Would
love to hear some thoughts on that and if there are any use cases that can
benefit from that.
> NM Container lifecycle and state transitions to support for PAUSED container
> state.
> -----------------------------------------------------------------------------------
>
> Key: YARN-5292
> URL: https://issues.apache.org/jira/browse/YARN-5292
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Hitesh Sharma
> Assignee: Hitesh Sharma
> Attachments: YARN-5292.001.patch, YARN-5292.002.patch,
> YARN-5292.003.patch, YARN-5292.004.patch, YARN-5292.005.patch, yarn-5292.pdf
>
>
> This JIRA addresses the NM Container and state machine and lifecycle changes
> needed to support pausing.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
