You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@yunikorn.apache.org by "Craig Condit (Jira)" <ji...@apache.org> on 2024/04/01 22:32:00 UTC

[jira] [Commented] (YUNIKORN-2517) [Yunikorn] Incorrect Placeholder Count for Duplicate Task Groups in Gang scheduling

    [ https://issues.apache.org/jira/browse/YUNIKORN-2517?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17832963#comment-17832963 ] 

Craig Condit commented on YUNIKORN-2517:
----------------------------------------

This is technically undefined behavior, as task groups need to be unique. I would submit that the current behavior is as good as any - we create the correct number of placeholders, with the specified resources. However, it's difficult to properly match a given pod with its placeholder if there are duplicate group names. We should probably keep the current behavior but log as WARN and also emit a pod event warning to let users know they have specified duplicate task groups.

> [Yunikorn] Incorrect Placeholder Count for Duplicate Task Groups in Gang scheduling
> -----------------------------------------------------------------------------------
>
>                 Key: YUNIKORN-2517
>                 URL: https://issues.apache.org/jira/browse/YUNIKORN-2517
>             Project: Apache YuniKorn
>          Issue Type: Bug
>          Components: core - scheduler
>            Reporter: Rajesh Kanhaiya Lal
>            Assignee: Manikandan R
>            Priority: Major
>
> Hi Team , I am getting incorrect placeholder count for the duplicate task groups in gang scheduling.
> Example :
> {code:java}
> TaskGroups: []v1alpha1.TaskGroup{
> 				{Name: "groupdup", MinMember: int32(3), MinResource: map[string]resource.Quantity{
> 					"cpu":    resource.MustParse("10m"),
> 					"memory": resource.MustParse("10M"),
> 				}},
> 				{Name: "groupdup", MinMember: int32(5), MinResource: map[string]resource.Quantity{
> 					"cpu":    resource.MustParse("10m"),
> 					"memory": resource.MustParse("10M"),
> 				}},
> 				{Name: "groupa", MinMember: int32(7), MinResource: map[string]resource.Quantity{
> 					"cpu":    resource.MustParse("10m"),
> 					"memory": resource.MustParse("10M"),
> 				}},
> 			}, {code}
> for the above config, we are getting a total of 17 pods ( 2 groups. + 15 Placeholders).
> It's adding the duplicate group placeholder as well.
> Could you please take a look?
> {code:java}
> gangjob-c805x-l4fx9                  1/1     Running   0          47s
> gangjob-c805x-tc8tr                  1/1     Running   0          47s
> tg-appid-oqina-groupa-1ap48pr4us     1/1     Running   0          45s
> tg-appid-oqina-groupa-25t5jubyzl     1/1     Running   0          45s
> tg-appid-oqina-groupa-6oxhqxnebc     1/1     Running   0          45s
> tg-appid-oqina-groupa-bqj9nk3mdq     1/1     Running   0          45s
> tg-appid-oqina-groupa-hugxbjb3xv     1/1     Running   0          45s
> tg-appid-oqina-groupa-o46k68fhw1     1/1     Running   0          45s
> tg-appid-oqina-groupa-vs5kxeop8z     1/1     Running   0          45s
> tg-appid-oqina-groupdup-786dl3gch2   1/1     Running   0          45s
> tg-appid-oqina-groupdup-877tnd4xdl   1/1     Running   0          45s
> tg-appid-oqina-groupdup-b7yef7w47x   1/1     Running   0          45s
> tg-appid-oqina-groupdup-cdqm1fcwbo   1/1     Running   0          45s
> tg-appid-oqina-groupdup-hlxwv9to9z   1/1     Running   0          45s
> tg-appid-oqina-groupdup-mvcd5pkijw   1/1     Running   0          45s
> tg-appid-oqina-groupdup-o4d9s8d02p   1/1     Running   0          45s
> tg-appid-oqina-groupdup-srrxrukstd   1/1     Running   0          45s {code}
>  
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

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