2016-07-12 18:46:16 |
Nate Potter |
description |
A volume that is in the 'error_extending' state isn't always deleted when it should be. I had five volumes, all in 'error_extending', and when I tried to delete them all only two were deleted. Here's the log:
http://paste.openstack.org/show/530768/ |
This seems like it's likely an issue for any driver rather than just LVM, but I have LVM enabled. I created a volume and snapshot, and then ended up manually deleting the snap with lv commands so that cinder didn't know about it. Then when trying to delete the snapshot with Cinder using the force param, it's impossible to remove the entry from Cinder's DB, because if a VolumeBackendApiException is raised when the snapshot isn't found, all that's done is set the snapshot's status to error_deleting. I think it might be useful to create a new exception for not finding the intended snapshot in the backend, and allow Cinder to delete it from the DB in this case. |
|