[
https://issues.apache.org/jira/browse/YUNIKORN-1562?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Craig Condit resolved YUNIKORN-1562.
------------------------------------
Fix Version/s: 1.3.0
Resolution: Fixed
Merged to master.
> plugin mode: remove status update completely
> --------------------------------------------
>
> Key: YUNIKORN-1562
> URL: https://issues.apache.org/jira/browse/YUNIKORN-1562
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: shim - kubernetes
> Reporter: Wilfred Spiegelenburg
> Assignee: Wilfred Spiegelenburg
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.3.0
>
>
> The k8shim in plugin mode only updates the pod status to move the pod from
> the _unschedulablePods_ structure to the activeQ. This caused an issue in the
> way we did it before YUNIKORN-1547 as we could leave a pod in a custom state
> that K8s does not know.
> After YUNIKORN-1547 we used the pending state. There could still be cases
> that the state set by the plugin showed up and not the state that the pod was
> really in. A pod could be fully scheduled and running while still showing
> pending as set by the plugin.
> The state change was added to the plugin code based on an incorrect
> assumption that it would trigger a move between the _unschedulablePods_ and
> {_}activeQ{_}. However the scheduler explicitly excluded pod status when
> checking for updates. We should thus not set the status of the pod to limit
> side effects and impact.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
