Description of the problem:

This week end i did the migration of my production from heroku to Railway. Over the past few days, our Gunicorn workers are consistently being killed with messages like:

Now I really have often this error in my backend web (django) Gunicorn workers killed by SIGKILL despite available memory (timeout issues) (never had this before)

Environment:

Railway service: Web (Django + Gunicorn)

Allocated memory: 4 GO

Configured workers: 3

Error message:

[CRITICAL] WORKER TIMEOUT (pid:…)

Worker (pid:…) was sent SIGKILL! Perhaps out of memory?

even though memory usage never exceeds ~1.8 GB out of our4GB allocation. We have already:

I Verified RAM usage—no spikes above 1.8 GB. (4GB allocation)

I Confirmed this is the Gunicorn timeout (default 30 s) and not the Linux OOM-killer. I Increased --timeout to 60 s in our startup script.

Despite these measures, workers continue to be killed after 60 s on certain, but my longest request took 3sec in heroku and here it shuts down after 60sec.

Could you check whether there’s any additional Railway-side limit (e.g. CPU throttling, I/O constraints) that might trigger these SIGKILLs?

Do you have best practices or recommendations for hosting Django/Gunicorn applications on Railway?

How can i fix it, i don't see what i can improve, i'll have to back to heroku it it continues like this

Thank you for your help!