[
https://issues.apache.org/jira/browse/YARN-2113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15944243#comment-15944243
]
Wangda Tan commented on YARN-2113:
----------------------------------
[~jlowe]/[~eepayne].
Discussed with [~sunilg]/[~vinodkv] offline. I think for different use cases,
they may prefer to treat precedence of user limit and priority differently. For
example, for some use cases they may want per-priority user-limit, so higher
priority apps can always get preferred. So this could be a policy-driven
behavior.
For now we're fine with what you suggested, which user limit takes precedence,
and once we have user-limit policy, we can make this to be a default policy.
And per_priority user_limit can be done in a much lower priority.
Also, discussed detailed implementations with [~sunilg], we came up with an
implementation which can address both of existing user-limit and per-priority
user-limit. Sunil will update the patch shortly.
Thanks,
> Add cross-user preemption within CapacityScheduler's leaf-queue
> ---------------------------------------------------------------
>
> Key: YARN-2113
> URL: https://issues.apache.org/jira/browse/YARN-2113
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: scheduler
> Reporter: Vinod Kumar Vavilapalli
> Assignee: Sunil G
> Attachments: YARN-2113.0001.patch, YARN-2113.0002.patch,
> YARN-2113.0003.patch, YARN-2113.v0.patch
>
>
> Preemption today only works across queues and moves around resources across
> queues per demand and usage. We should also have user-level preemption within
> a queue, to balance capacity across users in a predictable manner.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
