atomic-upgrade-svc

[cgwalters]

(Note: not called atomic-upgraded since the "d" there parses as English past tense, not Unix daemon)

Some deployments would like to enable automatic updates. We need a mode to make that work well.

See: #40
See: https://github.com/coreos/locksmith

[stefwalter]

Doesn't this daemon have to do all rpm-ostree operations (including rollback, upgrade, status, and probably compose, db) in order to provide the prerequisite locking and coordination? If so, why would it be called 'atomic-upgrade-svc'?

[cgwalters]

compose is currently a server side thing. db is read-only, and note because rpm-ostree never live-mutates the current system, it's completely safe to just read the rpm database at any time, whether via rpm-ostree db or just dropping down to rpm -qa.

So basically let's just focus on rollback, upgrade, and status?

[cgwalters]

This was done.

[cgwalters]

I think we still should have an example systemd unit timer for doing hands-off automatic updates for individual systems. That's an rpm-ostree issue which I'll file here. But at a cluster level one needs to script things more, there are a few projects/models around this like https://github.com/projectatomic/commissaire

[cgwalters]

Created #247 for the first part.

[ashcrow]

commissiare should already be able to do on demand upgrade and status. I think it would make sense to add rollback as well as an optional operation scheduler (or a controller of local operation schedulers).

[pgsamila]

Why rpm-ostree have an issue on automatic updates for individual system? If the system can updated manually, then the automatic is possible.

[cgwalters]

Let's talk in #247 for making it easier to turn on automatic host updates as well as scripting them.