Cron Service and API Service randomly crashing after no changes to them
idodekerobo
HOBBYOP

2 months ago

I have a cron worker and API service that is deployed via a monorepo that started crashing and not deploying after I pushed changes to the vercel hosted web app in the monorepo

I'm not sure why and will post as much information:

project id: c33d6859-c6e9-4fa3-bc67-6a74c5547e03

cron service start command: PYTHONPATH=src /opt/venv/bin/python -m app.workers.scheduler

api start command: PYTHONPATH=src /opt/venv/bin/uvicorn app.main:app --host 0.0.0.0 --port $PORT

can somone help me understand why out of no where my cron is crashing and my api isn't passing health checks?

i saw the outage related to metal based builds which seems to be related, but that now says its been resolved and the app still won't deploy properly.

attachments:

  • cron service deployment logs
  • api service crash logs
  • queue.py - the app.workers.queue file
  • health.py - health.py route
  • main.py - fastapi entrypoint
  • pyproject.toml - backend api imports
  • railway.toml - railway config
$20 Bounty

7 Replies

2 months ago

Have you tried using Railpack?


idodekerobo
HOBBYOP

2 months ago

i was coming here to say that i was using railpack, but after taking another look i think railway auto-turned on "Metal build environment" and added a deprecation flag to railpack.

i dont know much about metal build environment but it seems to be a new feature and related to the issues that were happening earlier?


2 months ago

Metal builds have been available for a year, I don't think Railway turned it on for you when it was off


2 months ago

Nixpacks indeed is deprecated in favour of Railpack


2 months ago

since q4 2025* oops


idodekerobo
HOBBYOP

2 months ago

nothing else changed in my config so i'm not sure what else the culprit can be. just turned off metal build environment and things seem to work and deploy now.

maybe its related to the earlier outage


ilyass012
FREE

2 months ago

looks like the metal build environment was the culprit here. it was causing the redis module to not be found at runtime even though it's correctly listed in your pyproject.toml. the fix is turning it off, and that lines up with the metal builder outage that was reported


Welcome!

Sign in to your Railway account to join the conversation.

Loading...