Today, rpm-ostree has base support for checking for updates, but not applying them: http://www.projectatomic.io/blog/2018/03/new-rpm-ostree-features/.
We should follow in the footsteps of CoreOS and turn on full automatic updates. For the cluster case, CoreOS has a tool called update-operator. We should investigate whether the operator pattern used there makes sense for Atomic Host, and if so, whether we can just adapt the update-operator codebase for Atomic Host.
I 💖 automatic updates.
Container Linux is relatively famous for doing auto-updates out of the box even on single node (non-Kubernetes) cases by default. Are we envisioning adopting that too? I would vote yes. Though it'd be nice to at least do some things like being less aggressive if an interactive shell is open or something?
Are we thinking this would also go across major versions? This goes back to https://pagure.io/atomic-wg/issue/231 etc.
It's also probably worth glancing at what Ubuntu Core does in this area too.
Metadata Update from @miabbott:
- Issue tagged with: meeting
+1 for automatic updates
This was proposed as an F29 change, though we expect F28 users to be able to turn this on in their clusters once it makes it into the TWRs. This also gives the opportunity to sync with the auto rollback GSOC work this planned for this summer.
+1 for automatic updates + reboots too
Metadata Update from @dustymabe:
- Issue untagged with: meeting
- Issue tagged with: host, question
+1 for automatic updates and reboots (scheduled or unattended) if possible.
I just got notified that I'm in at GSoC to work on the Atomic Host Upgrade Daemon =)
:clap: :confetti_ball: Awesome @lorbus!
to comment on this ticket.