cinder-volume-delete.txt

# Deleting an OpenStack Cinder Volume manually
#
# Only consider doing this when normal commands have failed
#
# https://ask.openstack.org/en/question/1378/cannot-remove-volume/
# https://raymii.org/s/articles/Fix_inconsistent_Openstack_volumes_and_instances_from_Cinder_and_Nova_via_the_database.html

Volume Deletion:
- Connect to the nova and cinder databases and update the device mappings there
   # VOL_ID=$(mysql -e "select id from nova.block_device_mapping where volume_id='<volume id>';"| egrep '[0-9]{1,5})
   # mysql -e "update nova.block_device_mapping set updated_at=now(),deleted_at=now(),deleted=$VOL_ID where volume_id='<volume id>';"
   # mysql -e "update cinder.volumes set updated_at=now(),deleted_at=now(),terminated_at=now(),mountpoint=NULL,instance_uuid=NULL,status='deleted',deleted=1 where deleted=0 and id='$volume_uuid';"
-ssh to the volume hosting node
-identify the target lun
   # tgt-admin -s | grep <volume-id>
-mark the target offline
   # tgt-admin --offline tid=<target id>
-get a list of all active connections to this lun (If there are none, skip the next step)
   # tgtadm --lld iscsi --op show --mode conn --tid <target id>
-close all of the active sessions (call this for each active session)
   # tgtadm --lld iscsi --op delete --mode conn --tid <target id> --sid <session id>
-delete the lun
   # tgtadm --lld iscsi --op delete --mode target --tid <target id>
-delete the target file in /var/lib/cinder/volumes
   # rm /var/lib/cinder/volumes/volume-<volume uuid>
-delete the logical volume
   # lvremove -f /path/to/volume
- Please note the above will leave a stale connection reference in the local iscsi connection table.  The connection itself is gone, however.  This doesn't cause any problems and can be corrected by running the following:
   # iscsiadm -m node -T <target name> -p <cinder host>:<port> -u