[
https://issues.apache.org/jira/browse/YUNIKORN-2271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17797321#comment-17797321
]
Weiwei Yang commented on YUNIKORN-2271:
---------------------------------------
Thanks a lot [~zhuqi] for looking into this!
> Incorrect handling of GPU only resources
> ----------------------------------------
>
> Key: YUNIKORN-2271
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2271
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: shim - kubernetes
> Reporter: Weiwei Yang
> Assignee: Qi Zhu
> Priority: Major
> Labels: pull-request-available
>
> https://github.com/apache/yunikorn-k8shim/blob/a118ba6c4d84804e2a407f9d91196ece4690cf09/pkg/common/resource.go#L61-L63
> this code seems to have a bug. When I define resource like this:
> {code}
> request:
> nvidia.com/gpu: 1
> limit:
> nvidia.com/gpu: 1
> {code}
> this is considered as QoS best effort and returned with just
> {code}
> Resources:
> pod:1
> {code}
> but I think this is a valid configuration that a pod only specifies GPU
> resource without memory or CPU. It seems this is the K8s upstream code:
> qos.GetPodQOS() causes this..
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
