Fastapi server did not survive a sprite restart

[khalido]

I tried out sprites.dev - very cool. To test sprite I made a small fastapi short story at https://wild-red-phoenix-jgh.sprites.app/ and shared the link. This worked great!

Next morning - the link was serving up "Bad Gateway"! 🙁 Not ideal obviously.

Claude Code says:

Service manager PID mismatch after VM boot causes bad gateway

After a VM restart, the Sprite service manager reports a stale PID for a running service, causing the HTTP proxy to return 502 Bad Gateway even though the underlying process is running and responding locally.

Steps to Reproduce:

1. Configure a service with http_port (e.g., FastAPI on port 8000)
2. Confirm service works externally
3. VM gets stopped/restarted (either manually or due to idle timeout)
4. Attempt to access the HTTP endpoint externally

Observed:

• External requests return 502 Bad Gateway
• curl localhost:8000 returns 200 (app is running)
• sprite-env services list shows a stale PID (e.g., 285) while actual process is PID 3
• Workaround: sprite-env services stop <name> && sprite-env services start ` fixes the issue

Suggested Fix: On VM boot, the service manager should either:

1. Reconcile PIDs with actual running processes before marking services as "running"
2. Or ensure the proxy health-checks the actual port before routing traffic