docker-compose.dev.yaml

# This is a development deployment for the application.  It has a single worker container and contains a definition for
# a redis server.  The Flask application is set to debug and the environment to development.
version: '3'

services:
  # This is the Flask application itself, which depends on both the redis server and the worker. There only needs
  # to be one of these containers deployed.  In order to share files with the worker(s), a common volume is mounted
  # to /working, which is where the Flask application will store files uploaded to it.
  web:
    build: .
    environment:
      - LC_ALL=C.UTF-8
      - LANG=C.UTF-8
      - REDIS_URL=redis://:@redis:6379/0
      - FLASK_ENV=development
      - DEBUG=True
      - SESSION_TTL_SEC=300
    ports:
      - "5000:5000"
    volumes:
      - latex-working:/working
      - ./:/var/www/app   # For local development
    depends_on:
      - redis
      - worker

  # This is the celery worker container, which is the same image but run with the COMPONENT environmental variable set
  # to 'worker' which causes the container to launch the celery worker process instead of the Flask application.  If
  # being deployed in swarm mode, 'replicas' can be used to set the number of copies, but otherwise you may simply need
  # to make multiple workers here if you want to run more than one worker per Flask app.  Multiple workers are likely
  # to only be necessary if there are many long compilation time sessions being performed.
  worker:
    build: .
    environment:
      - LC_ALL=C.UTF-8
      - LANG=C.UTF-8
      - REDIS_URL=redis://:@redis:6379/0
      - COMPONENT=worker
      - CELERY_LOG_LEVEL=DEBUG
    volumes:
      - latex-working:/working
      - ./:/var/www/app   # For local development
    depends_on:
      - redis

  # This is the celery beat scheduler container, which is also the same image but run with the COMPONENT environmental
  # variable set to 'scheduler' which causes the container to launch the celery beat process instead of the Flask
  # application.  There must only be one celery beat scheduler per deployment!
  beat:
    build: .
    environment:
      - LC_ALL=C.UTF-8
      - LANG=C.UTF-8
      - REDIS_URL=redis://:@redis:6379/0
      - COMPONENT=scheduler
      - CELERY_LOG_LEVEL=DEBUG
      - CLEAR_EXPIRED_INTERVAL_SEC=60
    depends_on:
      - redis

  # The Flask application and the workers communicate with each other via redis, and the Flask app uses redis to store
  # the non-file information uploaded by consumers of the service.  One redis server should be capable of servicing
  # many replicas of the compile service, so it is reasonable to use an external redis server in place of this one
  redis:
    image: redis:5

# The Flask application saves file data to a local folders in /working, so the workers need access to that data to
# perform the compilation task.  This is accomplished by having a volume shared by the two containers.
volumes:
  latex-working: