[
https://issues.apache.org/jira/browse/YARN-5216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15738826#comment-15738826
]
Hitesh Sharma commented on YARN-5216:
-------------------------------------
[~asuresh], I'm a little confused, so to be clear are you suggesting that we
have a preempt API in container executor and have a conf knob to select type of
preemption (PAUSE or KILL), is that correct?
Preempt is a very overloaded term and the current semantics work for scheduler,
but not sure whether they can be extended to work preservation, or some other
scenario. Currently if a container is preempted via PAUSE then it gets RESUMED
when there is free capacity but such a behavior may not be acceptable if the
scenario for calling preempt API is different. In other words, having the
scheduler call PAUSE or KILL is a very clear choice that needs to be made, but
we can't say that PAUSE should be the preemption policy for the container in
all cases. Further container PAUSE is only for opp. container and not something
we have enabled for GUARANTEED and thus I would scope it down to the scheduler
for now.
I do want to say that at some point in future it would make sense to have a
preempt API and a behavior that can be consistent across all scenarios, but I
don't think we are there yet, just my $0.02.
> 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: Sub-task
> Components: distributed-scheduling
> Reporter: Arun Suresh
> Assignee: Hitesh Sharma
> Labels: oct16-hard
> Attachments: YARN-5216-YARN-5972.001.patch, 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: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
