Summary

- Railway Postgres service is crash-looping and app deployment fails because DB is unreachable.

- Postgres logs show repeated recovery attempts followed by FATAL/PANIC due to disk full while writing WAL.

What I’m seeing

- Postgres startup:

- "database system was interrupted while in recovery"

- "automatic recovery in progress"

- redo completes, then:

- FATAL: could not write to file "pg_wal/xlogtemp.<pid>": No space left on device

- Postgres shuts down and repeats.

- This started after heavy app activity; project metrics show disk usage around ~494 MB for postgres-volume.

Relevant log excerpts

- 2026-02-09 02:17:26 UTC FATAL: could not write to file "pg_wal/xlogtemp.29": No space left on device

- Earlier also: PANIC: could not write to file "pg_wal/xlogtemp.4544": No space left on device

- Also saw errors writing tmp files: base/pgsql_tmp/... No space left on device

Impact

- Postgres never reaches “accepting connections”.

- App gudink-production) fails with ECONNREFUSED / failed to create database pool.

What I need help with

- How to increase/resize the Postgres disk/volume for this service (volume is named postgres-volume) OR the fastest recovery path (restore backup to a new Postgres instance with larger storage).

- If volume resize is not possible in my plan, confirm the correct migration/recovery steps.

Project details

- App repo: https://github.com/alejoras/gudink-app

- App already has an emergency kill switch to reduce load (DISABLE_BACKGROUND_JOBS=true) and won’t crash-loop when Postgres is down, but Postgres itself cannot boot until disk space is available.