Environment: Both backend (NestJS) and Postgres running as serverless (sleep when inactive)

Issue:

When incoming traffic triggers a cold start, both services are supposed to wake up together. However, intermittently the backend wakes up successfully but Postgres remains in a sleeping state and never comes online. The only way to recover is to manually trigger a redeploy on the Postgres service.

Expected behavior:

Both services should wake up reliably when traffic arrives, regardless of which one starts first.

Actual behavior:

Race condition on cold start — the backend initializes and attempts to connect to Postgres before it has finished waking up. Since there is no wake-up coordination between services, Postgres occasionally gets "stuck" in a sleeping state without completing its startup.

Frequency:

Intermittent — happens roughly 30–40% of cold starts.

Workaround:

Manual redeploy of the Postgres service restores normal operation.

Questions:

1. Is there a way to define a wake-up dependency order between services (e.g. backend waits for Postgres to be healthy before starting)?

2. Is this a known issue with serverless Postgres on Railway?

Thanks in advance.