Compose app containers restart because PostgreSQL authentication fails

The application started but did not answer on any port it publishes. The deploy log has the per-port result; a worker or queue-only app can ignore this. | The application is running but failing: it answered 502. Read the application log — the container is up, so this is the code inside it. | The application is restarting, not running. Docker reports windmill_server (Restarting (1) 48 seconds ago), windmill_worker (Restarting (1) 47 seconds ago), windmill_worker (Restarting (1) 47 seconds ago),…

Seen168 times · 8 variants

Why it happens

The Compose services start, but workers cannot authenticate to PostgreSQL with the configured postgres password. The affected containers restart, leaving Caddy unable to resolve windmill_server; the health check then returns 502. The rustls message appears incidental to the database failure.

What the log looks like

Deploy log
windmill_worker_native-1  | Connecting to database...windmill_worker_native-1  | Error: Connecting to database: error returned from database: password authentication failed for user "postgres"windmill_server           | Restarting (1)caddy-1                   | error: dial tcp: lookup windmill_server: no such hostHealth check: port 80 returned 502

Try PanelAlpha Engine yourself.

The easiest way to turn your server into a safe, controlled environment for your AI agent. Deploy, manage and maintain any project from Claude Code, Cursor, Codex or any other MCP client. Open source under Apache 2.0, free to run.

Requires Ubuntu 24.04 or 26.04 · 2 GB RAM · 1 CPURead the documentation