[
https://issues.apache.org/jira/browse/YUNIKORN-641?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17342357#comment-17342357
]
Wilfred Spiegelenburg commented on YUNIKORN-641:
------------------------------------------------
To clarify this a bit more in {{Application.timeoutPlaceholderProcessing()}} we
have Case 1 which cleans up the left over pods. For that case we need to
generate one, or more, events using the events plugin to communicate the
release of the placeholders to the application. Leveraging the event cache on
the core side we can send an event.
We can use an APP event. However we cannot hook it into the placeholder pod
that we are releasing as that placeholder is getting deleted. We should somehow
determine the "first pod" for the app and add the event(s) into that pod.
> Add events for placeholder timeout to pod
> -----------------------------------------
>
> Key: YUNIKORN-641
> URL: https://issues.apache.org/jira/browse/YUNIKORN-641
> Project: Apache YuniKorn
> Issue Type: Sub-task
> Components: core - scheduler, shim - kubernetes
> Reporter: Wilfred Spiegelenburg
> Assignee: Manikandan R
> Priority: Major
>
> If a placeholder times out we should communicate that fact with the pod that
> triggered the creation of the placeholders.
> An event should be added to the pod to show the timeout has occurred.
> Originally there could have been a request for 10 placeholders and only 5
> were used. The placeholder timeout events make it possible to track that on
> the original pod.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
