This directory provides a Docker Compose setup for running MLflow locally with a PostgreSQL backend store and MinIO (S3-compatible) artifact storage. It's intended for quick evaluation and local development.

• MLflow Tracking Server — exposed on your host (default http://localhost:5000).
• PostgreSQL — persists MLflow's metadata (experiments, runs, params, metrics).
• MinIO — stores run artifacts via an S3-compatible API.

Compose automatically reads configuration from a local .env file in this directory.

• Git
• Docker and Docker Compose
  • Windows/macOS: Docker Desktop
  • Linux: Docker Engine + the docker compose plugin

Verify your setup:

docker --version
docker compose version

git clone https://github.com/mlflow/mlflow.git
cd docker-compose

Copy the example environment file and modify as needed:

cp .env.dev.example .env

The .env file defines container image tags, ports, credentials, and storage configuration. Open it and review values before starting the stack.

Common variables :

• MLflow
  • MLFLOW_PORT=5000 — host port for the MLflow UI/API
  • MLFLOW_DEFAULT_ARTIFACT_ROOT=s3://mlflow/ — artifact store URI
  • MLFLOW_S3_ENDPOINT_URL=http://minio:9000 — S3 endpoint (inside the Compose network)
• PostgreSQL
  • POSTGRES_USER=mlflow
  • POSTGRES_PASSWORD=mlflow
  • POSTGRES_DB=mlflow
• MinIO (S3-compatible)
  • MINIO_ROOT_USER=minio
  • MINIO_ROOT_PASSWORD=minio123
  • MINIO_HOST=minio
  • MINIO_PORT=9000
  • MINIO_BUCKET=mlflow

docker compose up -d

This:

• Builds/pulls images as needed
• Creates a user-defined network
• Starts postgres, minio, and mlflow containers

Check status:

docker compose ps

View logs (useful on first run):

docker compose logs -f

Open the MLflow UI:

• URL: http://localhost:5000 (or the port set in .env)

You can now create experiments, run training scripts, and log metrics, parameters, and artifacts to this local MLflow instance.

To stop and remove the containers and network:

docker compose down

Data is preserved in Docker volumes. To remove volumes as well (irreversible), run:

docker compose down -v

• Verify connectivity
  If MLflow can't write artifacts, confirm your S3 settings:

  • MLFLOW_DEFAULT_ARTIFACT_ROOT points to your MinIO bucket (e.g., s3://mlflow/)
  • MLFLOW_S3_ENDPOINT_URL is reachable from the MLflow container (often http://minio:9000)

• Resetting the environment
  If you want a clean slate, stop the stack and remove volumes:

  docker compose down -v
  docker compose up -d

• Logs

  • MLflow server: docker compose logs -f mlflow
  • PostgreSQL: docker compose logs -f postgres
  • MinIO: docker compose logs -f minio

• Port conflicts
  If 5000 (or any other port) is in use, change it in .env and restart:

  docker compose down
  docker compose up -d

• MLflow uses PostgreSQL as the backend store for experiment/run metadata.
• MLflow uses MinIO as the artifact store via S3 APIs.
• Docker Compose wires services on a shared network; MLflow talks to PostgreSQL and MinIO by container name (e.g., postgres, minio).

• Point your training scripts to this server:

  export MLFLOW_TRACKING_URI=http://localhost:5000

• Start logging runs with mlflow.start_run() (Python) or the MLflow CLI.
• Customize the .env and docker-compose.yml to fit your local workflow (e.g., change image tags, add volumes, etc.).

You now have a fully local MLflow stack with persistent metadata and artifact storage—ideal for development and experimentation.