Unreasonable cinder container startup time when using NFS driver with cinder backup

[Will Szumski]

The kolla container seems to recursively set the permissions on all files in /var/lib/cinder. This directory is where the backup share gets mounted:

https://github.com/openstack/kolla-ansible/blob/a52cf61b2234d2f078dd2893dd37de63e20ea1aa/ansible/roles/cinder/templates/cinder.conf.j2#L44

Here is the troublesome bit of config:

https://github.com/openstack/kolla-ansible/blob/aa8b8798ac0eaec8293262706044c2b4aa304e44/ansible/roles/cinder/templates/cinder-volume.json.j2#L40

This can take a very long time if there are a large number of backups.

There is also an issue in the kolla container which does a similar recursive ownership change:

https://github.com/openstack/kolla/blob/1c5d98ecab0c1f3a4643698808a78ae449fe1f14/docker/cinder/cinder-volume/extend_start.sh#L4

I think the simplest solution is to change `backup_mount_point_base`. Thoughts?

[Mark Goddard (mgoddard)]

Those recursive perms changes always struck me as being a bit unnecessary. Surely anything created by the container should have the correct permissions? In some cases it has been required though.