Volume Snapshot Deletion Fail After Upgrade

Problem

• Deleting a snapshot of an existing volume ends in deleting state perpetually and never completes.

# openstack snapshot list --all

+--------------------------------------+------------------------+----------------------+-----------+---+| ID                                  | Name                  | Description          | Status    | Size |+--------------------------------------+------------------------+----------------------+-----------+-----+| 14b44a0d-25c9-4bc9-ac46-aec84e889f5a | btutil1-data-snapshot1 |                      | deleting  |  50 || 41af6331-091f-4d60-a623-4f5986836bed | btutil1-data-snapshot1 | foo snapshot        | deleting  |  50 || 6085eeaf-ce7c-4e79-b207-3f2b4bcd2b64 | volume_snapshot        | None                | available |  10 |+--------------------------------------+------------------------+----------------------+-----------+-----+|

• No activity can be seen in the Cinder volume logs /var/log/pf9/cindervolume-base.log.

Environment

• Platform9 Managed OpenStack - v3.6.0

Cause

• With PMO v3.6, the Cinder code base was upgraded to Pike. Since Ocata, the [DEFAULT] section with the driver-related options has no effect. Now all driver options are listed in [backend_defaults] section, that indicates that those options are effective only in this section and [[backend_name]] sections listed in enabled_backends
• As a result, the backend names get renamed after the upgrade. However, the new backend names are not updated for the existing volumes

Resolution

1. Update the existing volumes to point to the new backend.

# cinder-manage volume update_host --currenthost [old backend name] --newhost [new backend name]

2. Forcefully delete the snapshot

# openstack volume snapshot delete --force [snapshot UUID]

Additional Information

• PMO v3.6 Release Notes
• Cinder Pike Release Notes