Closed mruoss closed 2 months ago
This looks to be related to cluster-autoscaler/provisioningrequest.
Hi @mruoss, indeed Kueue is not responsible for satisfying the ProvisioningRequest.
In addition to the issue in cluster-autoscaler, I would suggest you reach out to your Google Cloud representative to file a feature request.
/close
@alculquicondor: Closing this issue.
What happened:
We are using Kueue with GKE's Dynamic Workload Scheduler. It works great but as soon as we try to attach a generic ephemeral volume to the pod, it won't get provisioned anymore.
Workload and a provisioning request are created. However, after a while, the provisioning request is marked as failed. Its status looks as follows. Note the error message
Provisioning Request's pods cannot be scheduled in the nodepool, affected nodepools: **REDACTED_LIST_OF_NODEPOOLS**
.What you expected to happen:
If I remove the volume or create a PVC and replace the volume declaration with something like the following, a node gets provisioned and the pod gets scheduled. I would expect the same behaviour for pods with generic ephemeral volumes.
How to reproduce it (as minimally and precisely as possible):
This is the Pod we're trying to get scheduled on a fresh
nvidia-l4
machine:Anything else we need to know?:
Environment:
kubectl version
):v1.29.4-gke.1043002
git describe --tags --dirty --always
):v0.6.2
cat /etc/os-release
):uname -a
):