[
https://issues.apache.org/jira/browse/YARN-2113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15951260#comment-15951260
]
Wangda Tan commented on YARN-2113:
----------------------------------
[~eepayne],
bq. With the latest patch, there is still a use case in which containers from
under-served users can be preempted. Please consider the following ..
In my opinion, I think this preemption is valid: once user submit app3 with
highest priority, it should get resource from other apps if user3's UL doesn't
meet. CS's behavior is also consistent to the preemption behavior: once the
preemption from app1/app2 happens, the resource goes to app3/users.
This is different from the example in your previous comment:
https://issues.apache.org/jira/browse/YARN-2113?focusedCommentId=15930339&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15930339.
In your previous example, preempted resource will go back to the app of the
preempted container, which means logics between scheduler/preemption is not
consistent.
> 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:
> TestNoIntraQueuePreemptionIfBelowUserLimitAndDifferentPrioritiesWithExtraUsers.txt,
> YARN-2113.0001.patch, YARN-2113.0002.patch, YARN-2113.0003.patch,
> YARN-2113.0004.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]
