Having trouble starting gunicorn with Django deployed from dockerfile resulting in 503

Trying to deploy a django project that was developed locally with this Docker File

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Install dependencies
COPY ./requirements.txt .
RUN pip install -r requirements.txt

# Copy project
COPY . .

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Install dependencies
COPY ./requirements.txt .
RUN pip install -r requirements.txt

# Copy project
COPY . .

From the Django Railway template I've set the railway.json to be:

{
    "$schema": "https://railway.app/railway.schema.json",
    "build": {
        "builder": "NIXPACKS"
    },
    "deploy": {
        "startCommand": "python manage.py migrate && gunicorn engi_alpha.wsgi",
        "restartPolicyType": "ON_FAILURE",
        "restartPolicyMaxRetries": 1
    }
}

{
    "$schema": "https://railway.app/railway.schema.json",
    "build": {
        "builder": "NIXPACKS"
    },
    "deploy": {
        "startCommand": "python manage.py migrate && gunicorn engi_alpha.wsgi",
        "restartPolicyType": "ON_FAILURE",
        "restartPolicyMaxRetries": 1
    }
}

I've also set ALLOWED_HOSTS to be *, though I would like to make that specific soon. After running railway up in a linked project the build appears to install all requirements fine, and proceeds to deploy. With a Postgres DB service I'm able to get the DB to run the migrations successfully, creating empty tables, however at that point the logs stop (no errors, just no more entries). I've attached a screenshot of the repo structure and the end of the railway up command. I feel like I'm missing something to do with either permissions, URLS or the deployment commands, but I've not got there yet. It's my first personal project in web dev so I;m sure the answer is obvious to many :p Any help appreciated 🙂

36 Replies

Percy•2y ago

Project ID: ee6fe531-fb8f-45f3-906e-f22374d487ac

zimapurpleOP•2y ago

ee6fe531-fb8f-45f3-906e-f22374d487ac

Brody•2y ago

does the logs in the railway UI also stop at this point?

zimapurpleOP•2y ago

They do

Brody•2y ago

looks like the migration runs, but it never gets to start gunicorn do you have a start setting configured somewhere else like in the service settings?

zimapurpleOP•2y ago

If you mean through the value in permissable-voleyball > settings > deploy > start command the value has a nite that say it's sourced from the settings file. Was that the question? Also, Ive just come back to the page and am now seeing Deployments Paused Builds need to drain. null

Brody•2y ago

yeah lets come back to this after the incidents are resolved https://discord.com/channels/713503345364697088/846875565357006878/1103346301829136414

zimapurpleOP•2y ago

Sure, thanks for taking a gander

Brody•2y ago

hey wanna try again?

zimapurpleOP•2y ago

Sure, I'll run a 'railway up' and see what happens.

zimapurpleOP•2y ago

Interesting, totally different error. Trying to work out if I changed something that would cause it. Looks like the behaviour is now to prioritise the settings file and not the docker file?

Brody•2y ago

click the code button and show me a screenshot

zimapurpleOP•2y ago

except it's also not accessing my settings file? locally it;s still got my start command

zimapurpleOP•2y ago

Brody•2y ago

you must not have your railway.json file in the right place

zimapurpleOP•2y ago

ohhhhh i;ve run the cli in the wrong place, was working in a diff project. 1 sec

Brody•2y ago

~~show me your dockerignore file~~ lol

zimapurpleOP•2y ago

XD thought it was odd I had to relink the project.... building (theres a fat dep tree here 😦 )

Brody•2y ago

whats your requirements.txt look like?

zimapurpleOP•2y ago

OK, build complete, succesful. Depployment completed, logs stop at no migrations to apply

asgiref==3.6.0
Bokeh == 3.1.0
crispy-bootstrap5==0.6
Django==4.0.10
django-allauth==0.50.0
django-crispy-forms==2.0
djlint==1.23.0
environs[django]==9.5.0
gunicorn==20.1.0
pandas==1.5.3
pandera==0.13.4
pytest-django==4.5.2
psycopg2-binary==2.9.3
requests==2.28.2
requests-mock==1.9.3
statsforecast==1.5.0
sqlparse==0.4.3
types-requests==2.28.11.17
whitenoise==6.2.0

asgiref==3.6.0
Bokeh == 3.1.0
crispy-bootstrap5==0.6
Django==4.0.10
django-allauth==0.50.0
django-crispy-forms==2.0
djlint==1.23.0
environs[django]==9.5.0
gunicorn==20.1.0
pandas==1.5.3
pandera==0.13.4
pytest-django==4.5.2
psycopg2-binary==2.9.3
requests==2.28.2
requests-mock==1.9.3
statsforecast==1.5.0
sqlparse==0.4.3
types-requests==2.28.11.17
whitenoise==6.2.0

statsforcast has a fat set of it;s own dependencies however, they all completed

Brody•2y ago

did build finish?

zimapurpleOP•2y ago

yup yup same behaviour

zimapurpleOP•2y ago

Is it possible the difference between my railway.json and the fact I have a dockerfile in the repo itself is causing issues? The deploy is obviously automatically pointing at my docker, event though the railway.json points it to "NIXPACKS"

Brody•2y ago

why not just remove the railway.json file and then do (hold on)

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Copy project
COPY . .

# Install dependencies
RUN pip install -r requirements.txt

# run migrations
RUN python manage.py migrate

# start server
CMD gunicorn engi_alpha.wsgi

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Copy project
COPY . .

# Install dependencies
RUN pip install -r requirements.txt

# run migrations
RUN python manage.py migrate

# start server
CMD gunicorn engi_alpha.wsgi

so yeah delete the railway.json file, and ust that as your dockerfile

zimapurpleOP•2y ago

I'll have a crack

zimapurpleOP•2y ago

Failed at RUN python manage.py migrate, e.g. the step before the logs dropped previously. Issue seems to be related to detecting the SECRET_KEY varable, which it must have found succesfully before and is set in teh env vars

Brody•2y ago

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Copy project
COPY . .

# Install dependencies
RUN pip install -r requirements.txt

# run migrations and start server
CMD python manage.py migrate && gunicorn engi_alpha.wsgi

# Pull base image
FROM python:3.10.4-slim-bullseye

# Set environment variables
ENV PIP_DISABLE_PIP_VERSION_CHECK 1
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1

# Set work directory
WORKDIR /code

# Copy project
COPY . .

# Install dependencies
RUN pip install -r requirements.txt

# run migrations and start server
CMD python manage.py migrate && gunicorn engi_alpha.wsgi

try that

zimapurpleOP•2y ago

well that built a lot faster, I guess railway caches the layer?

Brody•2y ago

maybe

zimapurpleOP•2y ago

OHO! Closer, got more logs. URL still errors, but 301 this time.

Brody•2y ago

okay and now its a code issue, that i unfortunately cant help with, off to google you go!

zimapurpleOP•2y ago

Sure, thanks for your help this evening, I'll head into the wild tomorrow.

Brody•2y ago

no problem!

zimapurpleOP•2y ago

Seemingly just solved using the same info as in this SO: https://stackoverflow.com/questions/55726610/django-ssl-redirection-on-heroku-too-many-redirects Specifically adding the line SECURE_PROXY_SSL_HEADER = ('HTTP_X_FORWARDED_PROTO', 'https') to the settings.py file. My current (and new) understanding is that this is a required setting when your project starts implementing more secure practices on proxy servers?

Stack Overflow

Django SSL redirection on Heroku: 'Too many redirects'

I have a web app deployed to Heroku with a custom domain name which DNS is managed through CloudFlare. What I want to do is redirect HTTP requests to HTTPS. After setting SECURE_SSL_REDIRECT to T...

zimapurpleOP•2y ago

@Brody I've spotted that railway has community templates, and there is one for Django, however I didn't start from there because a) I wasn't sure if this would ever be worth deploying, b) the template used is extremely basic, verging on unsafe unless configured? and c) I didn't actually know about railway in the first stage of the work. Therefore my question is effectively: Is this deliberately railways approach, e.g. ALL complexity of the framework is up to the user, or is there an interest in 1) extending the existing template or 2) creating a second "auth focussed" django template? (fine both ways btw, straight up honest question)

Brody•2y ago

the purpose of the framework templates (like django's template) is just to show the user how to get the project active on railway, the rest it completely up to the user

Gaming

Programming

Having trouble starting gunicorn with Django deployed from dockerfile resulting in 503