Description of problem: Volume and volumesnapshot did not be deleted in the back end after OCP cluster removal Version-Release number of selected component (if applicable): 4.7.0-0.nightly-2020-12-14-080124 How reproducible: Always Steps to Reproduce: 1. Install OSP cluster and cinder csi driver is installed. 2. Create pvc (reclaimPolicy: Delete in storageclass) 3. Create volumesnapshot (deletionPolicy: Delete in volumesnapshotclass) 4. Delete cluster 5. Check the volume and volumesnapshot in the back end Actual results: volume and volumesnapshot did not removed Expected results: volume and volumesnapshot should be removed after OCP cluster removal Master Log: Node Log (of failed PODs): PV Dump: PVC Dump: StorageClass Dump (if StorageClass used by PV/PVC): Additional info:
Cinder CSI does not offer the ability to set additional tags the resources it creates. It does add the "cinder.csi.openstack.org/cluster" property however [1], which I assume is to match the volume with a cluster ID, but in practice it is set to 'kubernetes': $ openstack volume show pvc-96cfac38-fa62-4a51-8f2d-febf96920b2a -f value -c properties {'cinder.csi.openstack.org/cluster': 'kubernetes'} The 'kubernetes' value apparently comes from [2]. Once we fix the cluster name for CSI driver, we'll have to adapt the installer's destroy module to look for volumes with the `cinder.csi.openstack.org/cluster=<cluster_id>` property. [1] https://github.com/openshift/cloud-provider-openstack/blob/ef0206889cb7c39f04d215e8cbb6c69616e59ea4/pkg/csi/cinder/controllerserver.go#L104 [2] https://github.com/openshift/openstack-cinder-csi-driver-operator/blob/729e71f/assets/controller.yaml#L49
*** Bug 1917710 has been marked as a duplicate of this bug. ***
Removing the Triaged keyword because: * the target release value is missing * the QE automation assessment (flag qe_test_coverage) is missing