Restart policies

WSLc has no restart policy of its own that covers every case (the wslc CLI accepts only no and always for --restart), so WSLc Desktop Manager adds an app-level restart policy.

It only runs while the app is running

The policy is enforced by WSLc Desktop Manager, not by WSLc. If the app is closed, a crashed container stays down until you open the app again. Turn on Start with Windows and Minimize to tray to keep the policy in effect all the time.

Policies

Policy	Restarts the container when…
Never (default)	Never.
Always	It stops for any reason, including a clean exit.
On failure	It exits with a non-zero code (an unknown exit code counts as a failure). Set Max restarts to cap the attempts.
Unless stopped	It stops for any reason, unless you stopped it from the app. A stop you made sticks across app restarts.

Choose a policy in the New container dialog or on the container details page. Policies are saved per container name in %LOCALAPPDATA%\WslcDesktop\restart-policies.json. Compose restart: values are applied to the services of a stack when you bring it up.

What happens

• When the app starts, containers whose policy says they should be running but aren't are started right away. A stop you made in an earlier app session only sticks for Unless stopped.
• While the app runs, a container that exits is restarted with a growing delay: 1, 2, 4, 8… seconds, up to 60 seconds.
• Stopping or deleting a container from the app never triggers a restart. Starting it again from the app clears any crash-loop state.
• A container that runs for 60 seconds without exiting is considered healthy and its attempt count resets.
• After 5 failed attempts in a row (or the Max restarts you set for On failure) the app stops trying and marks the container Crash loop. Fix the cause and start it yourself.
• Setting a policy on a container that is currently stopped doesn't start it.

The Restart column on the Containers page shows the policy and live status: Restarting in 4s, Crash loop, or the policy with a restart count.

How crashes are detected

The app watches the session's container list (every 2 seconds, 10 seconds while hidden in the tray) and reads the exit code from each container's status. This works the same whichever way the app is connected.

• App owns the session: the WSLc API's ProcessCrashed and Terminated events also trigger an immediate check, so restarts start sooner.
• App is attached to an existing WslcDesktop session (for example one left running by an earlier copy of the app): WSLc offers no way to subscribe to events from an attached client, so the app relies on polling alone. Restarts can lag by a few seconds.

Known gap

A container stopped outside the app (for example wslc stop in a terminal) looks like an exit to the app, so Always and On failure (when the stop gives a non-zero code) restart it. Stop containers from the app, or set the policy to Never first. If the whole session ends, the app can only restart containers once the session is back.

Native --restart

The app doesn't pass --restart to wslc. It is the single authority, which also makes on-failure and unless-stopped possible. If you create a container with wslc run --restart always yourself, WSLc restarts it as well; this is harmless because starting a running container does nothing.