docker-compose.yml

version: '3'

services:
  # Redis holds the data about compute-intensive jobs and is mainly used by Redis Queue (RD).
  redis:
    image: redis:alpine
    expose:
      - "6379"

  web-frontend:
    # The user-facing web frontend of Appcestry
    # Nginx hosts the static content of the UI and works as a reverse proxy for the HTTP-interface.
    image: jasonthc/appcestry-frontend:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    depends_on:
      - http-interface
    ports:
      - "80:80"
    expose:
      - "80"

  http-interface:
    # The HTTP-interface passes the data received from web-based UI to Appcestry.
    # For compute-intensive operations (AppGene conversion and comparison), tasks are stored as Redis Queue (RQ) jobs for rq-worker-* instances to take up.
    image: jasonthc/appcestry:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    depends_on:
      - redis
    expose:
      - "8899"
    entrypoint: /appcestry/app.py

  # For rq-worker-convert and rq-worker-compare:
  #  A rq-worker has to wait for Dask to complete a job before it can process a next one.
  #  Scaling a rq-worker to multiple instances is needed to handle multiple requests.

  rq-worker-convert:
    # Worker to pass an AppGene conversion job to dask-scheduler
    image: jasonthc/appcestry:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    depends_on:
      - http-interface
      - dask-scheduler
    entrypoint: /appcestry/rq_worker_convert.sh

  rq-worker-compare:
    # Worker to pass an AppGene comparison job to dask-scheduler
    image: jasonthc/appcestry:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    depends_on:
      - http-interface
      - dask-scheduler
    entrypoint: /appcestry/rq_worker_compare.sh

  dask-scheduler:
    # Assigning jobs to dask-worker(s) for distributed computing
    image: jasonthc/appcestry:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    expose:
      - "8786"
    command: /appcestry/dask_scheduler.sh

  dask-worker:
    # dask-worker executes tasks assigned by dask-scheduler.
    # On a single node, tasks are executed in multiple threads.
    # Or, to distribute the load to a cluster of machines, deploy instances of Dask worker on multiple nodes.
    image: jasonthc/appcestry:0.0.1b
    env_file:
      - ./appcestry_cluster.env
    depends_on:
      - dask-scheduler
    command: /appcestry/dask_worker.sh