[
https://issues.apache.org/jira/browse/YARN-5216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15374174#comment-15374174
]
Hitesh Sharma commented on YARN-5216:
-------------------------------------
We investigated a few approaches over here:
* Have a subclass of {{QueuingContainersManagerImpl}}: this approach has some
pros but the problem here is that subclassing just to override the preemption
behavior isn't the right thing to do.
* Having a pluggable policy in {{QueuingContainersManagerImpl}} requires
extension points to select which containers to run, run the container, preempt
the container, etc. This approach starts to get more complex as we look to add
support for PAUSED containers [YARN-5292].
Based on the feedback here and the discussions we have had, I'm looking into
adding support for PAUSED containers within {{QueuingContainersManagerImpl}}.
That would simplify things quite a bit and allow us to have a more pluggable
and cleaner design.
[~asuresh], [~kkaranasos], thanks for all the feedback!
> Expose configurable preemption policy for OPPORTUNISTIC containers running on
> the NM
> ------------------------------------------------------------------------------------
>
> Key: YARN-5216
> URL: https://issues.apache.org/jira/browse/YARN-5216
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Arun Suresh
> Assignee: Hitesh Sharma
> Fix For: 2.9.0
>
> Attachments: YARN5216.001.patch, yarn5216.002.patch
>
>
> Currently, the default action taken by the QueuingContainerManager,
> introduced in YARN-2883, when a GUARANTEED Container is scheduled on an NM
> with OPPORTUNISTIC containers using up resources, is to KILL the running
> OPPORTUNISTIC containers.
> This JIRA proposes to expose a configurable hook to allow the NM to take a
> different action.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
