[
https://issues.apache.org/jira/browse/YARN-5292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15720879#comment-15720879
]
Hitesh Sharma commented on YARN-5292:
-------------------------------------
Hi [~asuresh], thanks a lot for the feedback!
1. The default behavior is to throw an exception which is caught by the
ContainerLauncher and proceeds to kill the container. So if no PAUSE/RESUME
support exists then we kill the container. On a side note, we can open a JIRA
to implement PAUSE/RESUME for some of the executors like Docker.
2. Took care of collapsing transitions into one.
3. If the container is REINITIALIZLING and we get a PAUSE then the behavior is
undeterministic. Pausing the container when it hasn't finished reinitialization
can be be bad thus we kill instead. I feel it would be quite complicated if we
try to add the container back to the scheduler queue somehow thus let's not try
to do so.
4. Good point. Done.
Please have a look at the posted patch.
> Support for PAUSED container state
> ----------------------------------
>
> Key: YARN-5292
> URL: https://issues.apache.org/jira/browse/YARN-5292
> Project: Hadoop YARN
> Issue Type: New Feature
> 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.pdf
>
>
> YARN-2877 introduced OPPORTUNISTIC containers, and YARN-5216 proposes to add
> capability to customize how OPPORTUNISTIC containers get preempted.
> In this JIRA we propose introducing a PAUSED container state.
> When a running container gets preempted, it enters the PAUSED state, where it
> remains until resources get freed up on the node then the preempted container
> can resume to the running state.
>
> One scenario where this capability is useful is work preservation. How
> preemption is done, and whether the container supports it, is implementation
> specific.
> For instance, if the container is a virtual machine, then preempt would pause
> the VM and resume would restore it back to the running state.
> If the container doesn't support preemption, then preempt would default to
> killing the container.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
