Operating a box

Reboot & crash survival

• Docker installs run with --restart unless-stopped — Docker brings the container back after a reboot or crash.
• Native installs register a systemd service (Restart=always) when systemd is available, so the box survives reboots and restarts if a component dies. The launcher supervises all three processes (daemon, Caddy, agent) and exits as a unit if any one fails, so you never get a half-up box.
• Where systemd isn't available, native runs as a foreground process that does not survive a reboot — run motion start again after one.

Check which you have with motion status (it prints systemd or nohup).

Updating

motion update

Checks the hub's current version for your platform and updates in place if it's newer — downloading to a temp dir and atomically swapping only on success. motion doctor also surfaces "update available".

State (sessions, scheduler) is versioned: if a new build changes the on-disk format, the old state is archived and recreated rather than crash-looping — your files are never touched, only session/scheduler history.

Rotating the connect token

A connect token is a long-lived bearer credential. If one leaks, rotate it:

• In the dashboard, open the workspace's connect command and click Regenerate token. The old token stops working immediately, and the box must reconnect with the new one.

Tokens are 256-bit, compared in constant time, and ride an Authorization header (not the URL) so they stay out of access logs. On native installs the token is stored in ~/.motion/motion.env with 0600 permissions.

Reconnecting

If a box goes offline, just re-run the connect one-liner (or motion start if it's installed) — it stops any prior run and reconnects. The hub detects a dead box within ~50s via heartbeat, so the dashboard's online/offline status reflects reality.