You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@yunikorn.apache.org by "Weiwei Yang (Jira)" <ji...@apache.org> on 2020/04/10 17:40:00 UTC

[jira] [Commented] (YUNIKORN-88) Yunikorn webapp shows the apps as running even though they are completed and terminated

    [ https://issues.apache.org/jira/browse/YUNIKORN-88?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17080657#comment-17080657 ] 

Weiwei Yang commented on YUNIKORN-88:
-------------------------------------

Hi [~ayubpathan]

Thank you. This is a known issue.

Well, the reason is on K8s, it is "difficult" to tell if an app is completed. When we run workloads on K8s, yunikorn groups a bunch of pods and claim them belong to one application, usually leverage some meta info in the pods' spec, such as label "applicationID". YuniKorn only sees pods with app IDs,  we can tell if a pod is completed, but not the app. How can we know if there will be new pods added for this app?

This is something we need to improve for certain cases:
 # The framework already supports K8s operators, e.g spark-k8s-operator. The operator pattern manages the job's lifecycle in a CRD, so we can easily retrieve the status of a job from the CRD.
 # For the native Spark on k8s, with spark-submit. We can check the status of the spark driver pod, when the driver pod is finished, we can say the job is finished.

Unfortunately, we have to improve this case by case. There is no common fix for all sorts of applications so far I can see.

Please let me know your thoughts, thanks!

> Yunikorn webapp shows the apps as running even though they are completed and terminated
> ---------------------------------------------------------------------------------------
>
>                 Key: YUNIKORN-88
>                 URL: https://issues.apache.org/jira/browse/YUNIKORN-88
>             Project: Apache YuniKorn
>          Issue Type: Bug
>          Components: webapp
>            Reporter: Ayub Pathan
>            Assignee: Weiwei Yang
>            Priority: Major
>         Attachments: image-2020-04-09-19-18-03-414.png, image-2020-04-09-19-18-37-721.png
>
>
> Attaching the snapshots 
>  !image-2020-04-09-19-18-03-414.png|thumbnail! 
>  !image-2020-04-09-19-18-37-721.png|thumbnail! 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@yunikorn.apache.org
For additional commands, e-mail: issues-help@yunikorn.apache.org