a month ago
Hitting a strange issue - hoping someone has seen this before.
## The Problem
- Rollbacks work - app serves traffic, deploy logs appear
- Fresh builds fail - even the EXACT same commit that works as a rollback
- All requests get 502 "connection dial timeout"
- No deploy logs appear
## What I've Tried
- Reverted all code changes → still fails
- Cache-bust nixpacks.toml → still fails
- "Redeploy" on working commit → FAILS
- Rollback to same commit → WORKS
## Symptoms
- Build completes 
- Deploy completes (35-40s)
- Network shows "Not started"
- Zero deploy logs
- All 4 replicas timeout after 5s each
## Setup
- Nixpacks: Node 22, GDAL, awscli2
- Start: bash scripts/init-volume.sh && npm start
- Health check: /api/health
- 4 regions, V2 runtime
The working rollback also shows "Network: Not started" but works fine - so that's not the issue.
Has anyone seen cached rollbacks work but fresh builds fail? Stuck on old image and can't deploy updates. Any tips? 
4 Replies
a month ago
Try switching to a dockerfile instead of Nixpacks ...this gives you full control over the base image and eliminates railway's build environment as variable. Since rollback works, you know your app is fine ... issue is purely in how the new image is being built.
dharmateja
Try switching to a dockerfile instead of Nixpacks ...this gives you full control over the base image and eliminates railway's build environment as variable. Since rollback works, you know your app is fine ... issue is purely in how the new image is being built.
a month ago
Thanks! Good idea. Full control. I'll try this out now.
dharmateja
Try switching to a dockerfile instead of Nixpacks ...this gives you full control over the base image and eliminates railway's build environment as variable. Since rollback works, you know your app is fine ... issue is purely in how the new image is being built.
a month ago
UPDATE: Switched from Nixpacks to Dockerfile - same problem.
Build succeeds (159 seconds), deploy logs are empty, 502 on all requests.
This is confirmed to be a Railway infrastructure issue, not our code or build config:
- Nixpacks: fails
- Dockerfile: fails
- Rollback to cached image: works
Something in Railway's deploy pipeline is broken for fresh builds.
a month ago
I don't think this is related to the Feb 11 incident...
I have been deploying regularly throughout the last two days with no issues. Deployments suddenly stopped working at aournd 4pm EST today (Feb 13).
- Before around 4pm today: Deployments working fine
- After around 4pm today: ALL fresh builds fail, only cached rollbacks work
I changed nothing significant. Same codebase, same config. Tried:
- Nixpacks build: fails
- Dockerfile build: fails
- Rebuilding the EXACT commit that was working: fails
- Rollback to cached image: works
Something changed on Railway's infrastructure around 4pm EST today that broke fresh deployments for my project.
Project: topomiller (api.topomiller.com)
Status changed to Open brody • 28 days ago