Problem

When running a service, I currently have to manually check the endpoint repeatedly to determine if it’s ready to handle requests (e.g. the model is loaded) or if there are any health issues. This is particularly critical for production-grade services.

Solution

1. Add an option in the service configuration to define a URL that dstack can use to:
   a) Verify the service is up and healthy.
   b) Warm up the service.
2. If the service does not pass the health check within the configured timeout, mark it as unhealthy.

Workaround

No response

Would you like to help us implement this feature by sending a PR?

Yes