A scheduling deadlock between the Cluster Autoscaler and the local volume provisioner

[vladimir-rise]

When storageclass is configured with 'provisioner: kubernetes.io/no-provisioner' (default), GKE Cluster Autoscaler does not schedule a new node, hence no PVs are created, and the scheduled POD is in a pending state (node selector is to the new node only)

When storageclass is configured with 'provisioner: somthing-else', the GKE Cluster Autoscaler provides a new node, and new PVs are created by the local-volume-provisioner. However, a POD is still pending because PVC cannot be bound:

PVC description:
Waiting for a volume to be created either by the external provisioner 'somthing-else' or manually by the system administrator. If volume creation is delayed, please verify that the provisioner is running and correctly registered.

P.S.
Manual removing pending PVC causing it's recreation then binding occurs and POD is getting running