Unattended upgrades, hyper-converged units and Cloud upgrade via actions

[Alvaro Uria]

On smooshed applications (ie. ceph-osd, nova-compute),
0) set action-managed-upgrade=True
1) set ceph-osd openstack-origin="cloud:xenial-mitaka" and run the upgrade for the unit (via action)
2) Unattended upgrades runs in a daily cronjob

The above will cause upgrades of all the packages in the unit, including ie. nova-compute ones.

Fix:
1) Create an action to disable/enable unattended upgrades
2) Document a list of actions that need to be run to do an OpenStack upgrade (when action-managed-upgrade=True)

[Alvaro Uria (aluria)]

I've added the charms that are typically smooshed in the same machine.

I'm not sure if the fix should also be applied to other charms that generally run in containers (typically, the hacluster charm is a subordinate).

[James Page (james-page)]

Maybe the right thing todo is to disable unattended upgrades once a new origin is configured, and then re-enable them after the packages have actually been upgraded? This would avoid an out-of-charm control upgrade by the daily cronjob.

I'd prefer todo that rather than have users have to run actions todo the same.