[
https://issues.apache.org/jira/browse/YUNIKORN-2366?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Craig Condit updated YUNIKORN-2366:
-----------------------------------
Summary: Shim: Update GetPodResources() to handle in-place pod resource
updates (was: Update GetPodResources() to handle in-place pod resource updates)
> Shim: Update GetPodResources() to handle in-place pod resource updates
> ----------------------------------------------------------------------
>
> Key: YUNIKORN-2366
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2366
> Project: Apache YuniKorn
> Issue Type: Sub-task
> Components: shim - kubernetes
> Reporter: Craig Condit
> Assignee: Craig Condit
> Priority: Major
> Labels: pull-request-available
>
> The *GetPodResources()* function in the shim is used to convert a Pod's
> requested resources to a list of si.Resource objects. However, this is no
> longer sufficient once the *InPlacePodVerticalScaling* feature flag is
> enabled (or it becomes default in a future Kubernetes release).
> This function needs to also take into account the allocated resources in the
> {{Pod.Status.ContainerStatuses[].AllocatedResources}} collection in
> combination with {{Pod.Status.Resize}} to determine what the real resource
> utilization is. According to the
> [spec|https://github.com/kubernetes/enhancements/blob/master/keps/sig-node/1287-in-place-update-pod-resources/README.md],
> {{AllocatedResources}} should be used instead of
> {{Pod.Spec.Containers[].Resources.Requests}} unless it is not present. A
> special case also exists where Pod.Status.Resize is set to {{InProgress}} –
> in this case the larger of each resource should be used to avoid
> over-allocating resources on a node.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
