openmixerGitHub →

A poisoned autosave session

Symptom

The console comes back broken after every restart, in the same way each time. Restarting does not fix it. Sometimes nothing plays at all — every client connects, the surface draws, and no audio moves anywhere.

The tell is the repetition. A one-off failure is a fault; a failure that survives every restart is state being restored.

Why it happens

The server continuously writes a live autosave of the console, and on start it prefers that autosave over anything else. That is what makes a restart transparent during a show.

It also means a console left in a bad state writes that bad state to disk, and the next boot faithfully restores it. Two states are known to do real damage:

  • A persisted graph rate that flips the whole graph on boot — the console restores a rate the hardware cannot serve, and the audio device wedges.
  • A main output pinned to a device profile that is not active, or to a device that is no longer present — the mixer restores a route it cannot lay.

A wedged driver is not a quiet failure: with the graph's driver stalled, no client can play, not just the mixer.

Fix

Move the live autosave aside and boot clean. Do not delete it — it is evidence, and it may contain a show you want to salvage.

systemctl --user stop openmixer-server

cd ~/.local/state/openmixer/sessions          # or $XDG_STATE_HOME/openmixer/sessions
ls -la
mv autosave-live.json  autosave-live.broken-$(date +%F).json

systemctl --user start openmixer-server

The exact filename follows the reserved id autosave-live; list the directory rather than guessing. If the fault persists, move autosave-preload aside as well.

On the next start the server finds no live autosave, falls back to the most recent named session, and if there is none, boots the configured rig fresh.

Then rebuild deliberately

Load a named session — one you saved yourself, not an autosave — and check it before you trust it:

  • The graph rate, in the telemetry panel. It should be what the rig runs at, normally 48000 with a stagebox.
  • The main output route, in the patchbay's output routing row.
  • That sound actually reaches the speakers.

Save it under a new name once it is right. That gives the server a clean named session to fall back to next time.

Preventing it

  • Save a known-good named session for every rig, and re-save it whenever the rig changes. The fallback chain — live autosave, then the most recent named session, then a fresh rig — is only as good as the named session behind it.
  • Take a backup before an upgrade or a rig change. See backing up sessions.
  • Do not leave the console in a broken state at the end of a night. The autosave will remember it.

Related