c5ad221b-8b00-4b24-b8c9-eac5b00a9054

If your project isn’t detected as a python project (doesn’t have a main.py or isn’t the primary language) then you’ll have to tell nixpacks to install python. Create a nixpacks.toml file and add “python3” to the providers section

https://nixpacks.com/docs/configuration/file

make sure to include the “…” as well as that tells nixpacks to install the detected packages as well

@Charles_ for visibility ^

the real question is, why are you having your build clone a repo instead of letting railway's infra do that for you?

it seems that adding requirements.txt at project root fixed it - but I only need setup.py. Would that also work with nixpacks.toml if I remove requirements.txt?

@Brody would love to hear the better alternative - this does feel clunky

I'm actually installing 6 different private module which makes for a very long and awkward buildCommand

attach your repo to the service

That's what I've done for the main app, but it has deps on 6 private modules

Fair, this sounds like an absolute pain to do with nixpacks, i would highly recommend this be done with a Dockerfile

Thanks I will give it a go

@Brody how do you correctly pass ENV's definee in Railway into Dockerfile?

@Brody any idea why I can't access my application via url provided by Railway?

I see this in logs

2024-04-11 12:01:18 +0000 - dagster-webserver - INFO - Serving dagster-webserver on http://0.0.0.0:3000 in process 8

I have port defined in my variables

PORT=3000

I expose port in Dockfile:

EXPOSE 3000

my supervisord.conf

[supervisord]
nodaemon=true
user=root

[program:dagster-daemon]
command=dagster-daemon run
autostart=true
autorestart=true
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=0
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=0
environment=
    DAGSTER_HOME="/app"

[program:dagit]
command=dagster-webserver -h  0.0.0.0 -p %(ENV_PORT)s
directory=/app
autostart=true
autorestart=true
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes=0
stderr_logfile=/dev/stderr
stderr_logfile_maxbytes=0
environment=
    DAGSTER_HOME="/app"

@Brody just checking; are you aware of any issues/downsides from using supervisord inside Dockerfile for managing long-running processes?

I would have them split as two railway services but both processes need access to a shared volume which AFAIK isn't possible.

I use supervisor on a demanding php-fpm service on Railway and never had any issues with it.

nice, good to hear!

yeah it's totally fine, but personally, I much prefer parallel

far far simpler

Thanks, how can I avoid this Dockerfile caching? This layers runs git clone & pip install on repo's I've made commits to

set a service variable NO_CACHE to 1

@Brody please send me friend request if you're interested in paid work (deploying docker-compose to Railway)

well I'm honestly just more curious as to what you want to deploy more than anything