[
https://issues.apache.org/jira/browse/YUNIKORN-1089?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17500690#comment-17500690
]
Chung En Lee commented on YUNIKORN-1089:
----------------------------------------
I have another found about gang scheduler. I submit a spark application. If
driver pod has a _task-group-name_ that is not defined in _task-groups_ and
executor pods defined correctly. The placeholder pod still run until
placeholder timeout, after that they terminate. However, the executer place
holder stiil occupy the resource from the restful APIs.
> Application handling with invalid task group annotations
> --------------------------------------------------------
>
> Key: YUNIKORN-1089
> URL: https://issues.apache.org/jira/browse/YUNIKORN-1089
> Project: Apache YuniKorn
> Issue Type: New Feature
> Components: shim - kubernetes
> Reporter: Wilfred Spiegelenburg
> Priority: Major
>
> In YUNIKORN-1076 we added writing an event to the pod if the task group
> annotation parsing for a gang request failed. We currently accept the
> application and schedule it as if it does not have a gang request defined.
> This might not be what the end user expects.
> The Jira also mentioned failing or rejecting the application if the parsing
> failed. This is a major change:
> * the shim currently does never reject an application
> * is a broken annotation enough reason to reject the application
> Moved out the reject or more invasive handling out of the original Jira so we
> can look at it in more detail.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
