[
https://issues.apache.org/jira/browse/YUNIKORN-676?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17363234#comment-17363234
]
Weiwei Yang commented on YUNIKORN-676:
--------------------------------------
Thanks [~yuchaoran2011], I think we can close this. And for GS cases, we might
want to disable reservations by default. cc [~wilfreds].
Let me create a separate issue to track the discussion, thanks!
> App pending in one queue seems to block app scheduling on the other queue
> -------------------------------------------------------------------------
>
> Key: YUNIKORN-676
> URL: https://issues.apache.org/jira/browse/YUNIKORN-676
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: core - scheduler, shim - kubernetes
> Reporter: Weiwei Yang
> Priority: Blocker
> Fix For: 0.11
>
> Attachments: queues.yaml, s1.yaml, s2.yaml, s3.yaml
>
>
> [~yuchaoran2011] found this issue on their cluster.
> It looks like when we have 2 queues, root.a and root.b, if we submit an app
> with invalid node constraints (e.g node affinity to some non-exist node) to
> root.a, the app will be pending (expected); however, in some cases, apps in
> root.b are blocked. Restarting the scheduler seems to be able to solve this.
> Root cause unclear yet, reproduction steps unclear as well at the moment.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
