issue Dockerise Django Cuda application using docker compose

Question

I am trying to dockerize a Django Cuda application that runs on Nginx and Gunicorn.Problem is when I go to do prediction .. I get an error cuda drivers not found

My DockerFile:

FROM nvidia/cuda

FROM python:3.6.8
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1
WORKDIR /app
COPY ./requirements.txt /app/requirements.txt
RUN python -m pip install --upgrade pip
RUN pip install cmake
RUN pip install opencv-python==4.2.0.32
# RUN pip install pywin32==227
RUN pip install -r requirements.txt
COPY . /app
RUN python manage.py collectstatic --noinput
RUN pip install gunicorn
RUN mkdir -p /home/app/staticfiles/

Ngnix DockerFile

FROM nginx:1.21-alpine
RUN rm /etc/nginx/conf.d/default.conf
COPY nginx.conf /etc/nginx/conf.d

Ngnix config file

upstream project_settings {
    server web:8000;
}

server {

    listen 80;
    client_max_body_size 0;

    location / {
        proxy_pass http://project_settings;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Host $host;
        proxy_redirect off;
    }
    location /static/ {
        alias /home/app/staticfiles/;
    }
}

Main Docker compose file

services:
  nginx:
    build: ./nginx
    ports:
      - 1300:80
    volumes:
      - static_volume:/home/app/staticfiles/
    depends_on:
      - web
  web:
    build: .
    command: gunicorn project_settings.wsgi:application --bind 0.0.0.0:8000
    volumes:
      - static_volume:/home/app/staticfiles/
    image: sampleapp1121asa
    expose:
      - 8000
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [ gpu ]
volumes:
  static_volume:

Things are not working with docker compose, when I try to build the dockerfile seperately and then run using docker run --rm --gpus all -p 8000:8000 deefakedetectiondockerimage python3 manage.py runserver 0.0.0.0:8000 it works but the problem with this approach is I can not serve static file in docker. Ngnix is required to serve the static file, it means I need to run this through docker compose only

Answer 1

I found the solution for the same. Actually, things become difficult to run from docker-compose when you are trying to run multiple images in single container.

So I build the image using DockerFile for application and separate image for Ngnix and enable communication of both the container with unix socket connections.

My updated dockerfile for application:

#pull the nvidia cuda GPU docker image
FROM nvidia/cuda

#pull python 3.6.8 docker image
FROM python:3.6.8
ENV PYTHONDONTWRITEBYTECODE 1
ENV PYTHONUNBUFFERED 1
#create a directory to serve static files 
RUN mkdir -p /home/app/staticfiles/app/uploaded_videos/
WORKDIR /app
COPY ./requirements.txt /app/requirements.txt
RUN python -m pip install --upgrade pip
RUN pip install cmake
RUN pip install opencv-python==4.2.0.32
RUN pip install -r requirements.txt
COPY . /app
RUN python manage.py collectstatic --noinput
RUN pip install gunicorn
RUN mkdir -p /app/uploaded_videos/app/uploaded_videos/

VOLUME /app/run/
ENTRYPOINT ["/app/bin/gunicorn_start.sh"]

gunicorn_start.sh script

#!/bin/bash

NAME="project_settings"                                  # Name of the application
DJANGODIR=/app             # Django project directory
SOCKFILE=/app/run/gunicorn.sock  # we will communicte using this unix socket
NUM_WORKERS=3                                     # how many worker processes should Gunicorn spawn
DJANGO_SETTINGS_MODULE=project_settings.settings             # which settings file should Django use
DJANGO_WSGI_MODULE=project_settings.wsgi                     # WSGI module name

echo "Starting $NAME as `whoami`"

# Create the run directory if it doesn't exist
RUNDIR=$(dirname $SOCKFILE)
test -d $RUNDIR || mkdir -p $RUNDIR

# Start your Django Gunicorn
 gunicorn project_settings.wsgi:application --bind=unix:$SOCKFILE --workers $NUM_WORKERS --timeout 600

My updated docker file for Nginx

FROM nginx
WORKDIR /etc/nginx/
RUN rm /etc/nginx/conf.d/default.conf
COPY nginx.conf /etc/nginx/conf.d
EXPOSE 80

For step by step process you can follow this blog