docker-compose.yml

version: '3.3'

services:
  batch_prediction:
    build: ./prediction_service_batch/
    image: prediction_service_batch
    container_name: prediction_service_batch
    volumes:
      - ./data:/home/ubuntu/mlops_zoomcamp_homework/data
    networks:
      - frontend
    environment:
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - AWS_DEFAULT_REGION=${AWS_DEFAULT_REGION}
      - TRACKING_SERVER_HOST=${TRACKING_SERVER_HOST}
      - MLFLOW_BUCKET_NAME=${MLFLOW_BUCKET_NAME}
    depends_on:
      - web
      - nginx

  prediction_service:
    build: ./prediction_service_stream/
    image: prediction_service_stream
    container_name: prediction_service_stream
    depends_on:
      - evidently_service
      - mongo
      - web
    env_file:
      - .env
    environment:
      TRACKING_SERVER_HOST: "mlflow_nginx"
      EVIDENTLY_SERVICE_ADDRESS: "http://evidently_service:8085"
      MONGODB_ADDRESS: "mongodb://mongo:27017/"
    ports:
      - "9696:9696"
    networks:
      - backend
      - frontend

  web:
    restart: always
    build: ./mlflow
    image: mlflow_server
    container_name: mlflow_server
    expose:
      - "5000"
    networks:
      - frontend
      - backend
    environment:
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - AWS_DEFAULT_REGION=${AWS_DEFAULT_REGION}
    command: mlflow server --backend-store-uri postgresql://${POSTGRES_USER}:${POSTGRES_PASSWORD}@${POSTGRES_HOST}:5432/${POSTGRES_DATABASE} --default-artifact-root s3://${MLFLOW_BUCKET_NAME} --host 0.0.0.0 -p 5000

  nginx:
    restart: always
    build: ./nginx
    image: mlflow_nginx
    container_name: mlflow_nginx
    ports:
      - "80:80"
    networks:
      - frontend
    depends_on:
      - web

  mongo:
    image: mongo
    container_name: mongo
    ports:
      - 27017:27017
    volumes:
      - mongo_data:/data/db
    networks:
      - backend
      - frontend

  prometheus:
    image: prom/prometheus
    container_name: prometheus
    volumes:
      - ./05-monitoring/evidently_service/config/prometheus.yml:/etc/prometheus/prometheus.yml
      - prometheus_data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'
    ports:
      - "9091:9090"
    networks:
      - backend
    restart: always

  grafana:
    image: grafana/grafana
    container_name: grafana
    user: "472"
    depends_on:
      - prometheus
    ports:
      - "3000:3000"
    volumes:
      - ./05-monitoring/evidently_service/config/grafana_datasources.yaml:/etc/grafana/provisioning/datasources/datasource.yaml:ro
      - ./05-monitoring/evidently_service/config/grafana_dashboards.yaml:/etc/grafana/provisioning/dashboards/dashboards.yaml:ro
      - ./05-monitoring/evidently_service/dashboards:/opt/grafana/dashboards
      - grafana_data:/var/lib/grafana
    networks:
      - backend
      - frontend
    restart: always

  evidently_service:
    build:
      context: ./05-monitoring/evidently_service
      dockerfile: Dockerfile
    image: evidently_service
    container_name: evidently_service
    depends_on:
      - grafana
    volumes:
      - ./05-monitoring/evidently_service/datasets:/app/datasets
      - ./05-monitoring/evidently_service/config.yaml:/app/config.yaml
    ports:
      - "8085:8085"
    networks:
      - backend
      - frontend

networks:
  frontend:
    driver: bridge
  backend:
    driver: bridge

volumes:
  grafana_data: {}
  prometheus_data: {}
  mongo_data: {}

# based on https://towardsdatascience.com/deploy-mlflow-with-docker-compose-8059f16b6039
# but used postgresql db instead no aws as done in mlops zoomcamp course