Upgrading

`dnf upgrade` does not restart the mixer

This is deliberate, and it is the single most important thing to know about upgrading a live rig.

sudo dnf upgrade openmixer openmixer-server openmixer-web-ui

replaces the files on disk and the unit definitions, but the running engine keeps serving the pre-upgrade code. The package's uninstall scriptlets use the plain systemd user macros, not the restart-on-upgrade variants. A console mid-show is never bounced by a package transaction.

Nothing changes until you restart the units yourself.

Restarting after an upgrade

Pick a moment when silence is acceptable, then:

systemctl --user restart openmixer-server
systemctl --user restart reac-pw     # if this rig drives a stagebox

The source tree also carries scripts/deploy-live.sh, which gates the restart behind a shadow boot of the newly-installed server against a copy of the rig's real session on a spare port, a reac-pw capability check, and a post-restart health check. If you have the checkout, prefer it — it catches an upgrade that cannot load this rig's session before the live one goes down.

Session and scene files across versions

Sessions, scenes, patches and channel configs are JSON files under the state directory (see state and session directories). They are written by the server and read back by the server; a version change does not rewrite them until you save again.

Before an upgrade on a rig whose sessions matter, copy the state directory:

cp -a ~/.local/state/openmixer ~/openmixer-state-$(date +%F)

That is the whole backup. Restoring is a copy back with the units stopped.

Configuration files are preserved

/etc/openmixer/config.json is %config(noreplace): your edited copy survives, and the package's new version lands beside it as .rpmnew if the shipped default changed.
~/.config/reac-pw/ is written by the mixer's own Setup screen and is not owned by any package, so no upgrade touches it.
The nginx drop-in /etc/nginx/conf.d/openmixer-web-ui.conf is package-owned. If you edited the listen port or the upstream by hand, keep a copy — and reload nginx after an upgrade:
```
sudo systemctl reload nginx
```

After restarting, verify

systemctl --user status openmixer-server reac-pw
curl -s http://127.0.0.1:8080/health
getcap /usr/bin/reac-pw          # a reac-pw upgrade that lost its caps shows up here

Then load your session and check that the sources you expect are present. A session records the hardware it expected when it was saved, so a restore onto a rig missing a box tells you what is absent instead of failing quietly.

Downgrading

dnf downgrade on the same package set works and, like an upgrade, changes nothing running until you restart. Sessions saved by a newer server are not guaranteed to load into an older one — this is what the pre-upgrade copy of the state directory is for.