Workspace: Hagatun IT

Project: Fleet Analyzer

Project ID: 29384a88-9816-4694-b661-948a68e70025

Environment: production (64399704-8a81-4d68-a388-132a4cfc213c)

Services:

• Dashboard: 2858b60e-f607-47bd-b980-d93db019ad80
• Postgres: 6dddc998-fff8-40b3-9d19-e06e98fdf327
• Worker: 0f29e0df-fa0a-44be-afff-8532f619b836

Related / same symptoms

https://station.railway.com/questions/private-networking-service-cannot-reach-3d1be833

Same pattern reported there: private host times out, DATABASE_PUBLIC_URL works, Postgres itself is healthy.

Summary

After a Postgres restart / volume remount, services in the same project/environment cannot connect to Postgres over private networking. The public TCP proxy works. Production is currently mitigated by pointing Dashboard at DATABASE_PUBLIC_URL.

Timeline (UTC)

• ~2026-07-18 17:15: Postgres logs stop normal activity and show:

  "Mounting volume on: .../vol_v49eqi9tdokti534"

• 2026-07-19 ~12:36+: production site down; Dashboard pre-deploy

  pnpm run --filter @btr/core db:migrate exits 1 with little/no migrate output

• After Postgres restart, Dashboard starts and Redis connects over private networking, but Postgres fails repeatedly with:

  "PgClient: Connection timed out" / "Connection terminated due to connection timeout"

• Healthcheck /resources/healthcheck fails because it requires DB connectivity

Reproduction / evidence

1. Dashboard DATABASE_URL was (and is normally):

   ${{Postgres.DATABASE_URL}} → host postgres.railway.internal:5432

   Result: connect timeouts from Dashboard container. No useful auth/connection errors in Postgres logs during those attempts.

2. From my laptop against public proxy:

   psql -h tramway.proxy.rlwy.net -p 44024 -U postgres -d railway

   SELECT 1; → success (SSL works)

3. Workaround that restored production:

   Dashboard DATABASE_URL = ${{Postgres.DATABASE_PUBLIC_URL}}

   Redeploy → migrate + healthchecks succeed, site up.

4. Switching back to ${{Postgres.DATABASE_URL}} fails again with the same private-network timeouts.

Notes

• Dashboard and Postgres are in the same project and environment (production).
• Redis private connectivity from Dashboard continued to work while Postgres private did not.
• This does not look application-level: same app works immediately when only the DB URL host changes from internal → public proxy.

Ask

Please investigate private networking / DNS / routing between Dashboard and Postgres in this environment (same class of issue as thread 3d1be833). Public proxy works; private host postgres.railway.internal does not from the Dashboard service after the volume remount/restart.

Happy to provide additional deploy IDs, metrics screenshots, or a temporary shell if needed.