Docker

For production usage, use the docker build and a persistent cache and database (aka redis and postgres).

Read more about services overview

This allows to deploy the prefect server (and ui) seperated from one or more agents. Server and agent use the same image but with a different entrypoint.

investigraph docker image

Environment

Example settings for production, using postgres and redis:

PREFECT_API_URL=http://server:4200/api
PREFECT_API_DATABASE_CONNECTION_URL=postgresql+asyncpg://investigraph:investigraph@postgres/investigraph
FTM_STORE_URI=postgresql://investigraph:investigraph@postgres/investigraph
REDIS_URL=redis://redis:6379

Store this as a file to use as env_file within compose.

Example docker-compose.yml

This is a minimal production example using dedicated redis cache and persistent postgres database. Note that (as in the environment settings above) ftmstore and prefect.io share the same database instance.

Scale worker agents via

docker compose up --scale worker=4

Note that if you want to output result data to a file directory that should be shared (and visible from outside docker), add a volume mount to the agent service.

version: "3.9"

services:
  redis:
    image: redis:latest

  postgres:
    image: postgres:15
    volumes:
      - db:/var/lib/postgresql
    environment:
      POSTGRES_PASSWORD: investigraph
      POSTGRES_USER: investigraph
      POSTGRES_DB: investigraph

  server:
    image: ghcr.io/investigativedata/investigraph:latest
    ports:
      - 127.0.0.1:4200:4200
    depends_on:
      - postgres
      - redis
    volumes:
      - data:/data/prefect
      - ./datasets:/data/local/datasets
    env_file:
      investigraph.env

  agent:
    image: ghcr.io/investigativedata/investigraph:latest
    command: prefect agent start -q 'default'
    depends_on:
      - server
    env_file:
      investigraph.env

volumes:
  db:
  data: