Our web service repeatedly fails to connect to the Postgres instance. After adding the dashboard “arrow” link to Postgres-20i6, deployments still fail almost immediately:

ERROR:api.auth:Signup error: connection to server at "postgres.railway.internal" (...), port xxxx failed: Connection refused

[2025-11-11 23:11:22 +0000] [1] [ERROR] Worker (pid:4) was sent SIGKILL! Perhaps out of memory?

The app should be using:

postgresql://postgres:mRSlyFdsqjbsbLbWNqBOqOMWvwJBKROV@postgres-20i6.railway.internal:5432/railway

When we deploy from /backend, Railway builds successfully, but shortly after boot the workers are terminated and /api/auth/signup times out. Direct DB access through the proxy URL (shinkansen.proxy.rlwy.net:58104) works and schema is initialized (6 tables), so Postgres itself is up.

Steps taken:

1. railway variables --set confirmed DATABASE_URL is correct.

1. Re-linked web to Postgres-20i6 via dashboard (arrow now shows).

1. Redeployed (railway up & railway redeploy from /backend).

1. Initialized schema via proxy (init_railway_schema.py).

1. Smoke tests still fail; logs continue to show connection refused on the internal hostname.

Impact: /api/auth/signup and /api/auth/login return 500/timeout—auth is down in production.

Could you check the service link or networking between these services? Let me know if you need more logs or a reproduction.

Thanks!