[
https://issues.apache.org/jira/browse/YUNIKORN-84?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17080660#comment-17080660
]
Weiwei Yang commented on YUNIKORN-84:
-------------------------------------
Covering this case makes sense. This can prevent certain race conditions. Such
as during recovery, admin removes a node; or during the recovery, an app is
removed. All such cases will cause a node to be rejected by the scheduler.
I think we need more testing for this, run some chaos monkey scenarios to
ensure our recovery logic is robust enough.
> Always waiting in recovery phase as long as rejected node exists
> ----------------------------------------------------------------
>
> Key: YUNIKORN-84
> URL: https://issues.apache.org/jira/browse/YUNIKORN-84
> Project: Apache YuniKorn
> Issue Type: Bug
> Reporter: Tao Yang
> Priority: Major
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> In recovery phase, scheduler keeps waiting until all nodes are recovered, but
> not all nodes are considered according to Context#recover in
> context_recovery.go, only nodes with Healthy or Draining state will be
> counted, nodes with Rejected state should be counted as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]