MLFlow Image: How to make MLflow stateful while using in docker-compose

[Ikram Ul Haq]

Hey,
I am using the MLFlow Image provided here(https://hub.docker.com/r/ubuntu/mlflow) in one of the projects to store models and record runs. I have mounted a volume as well. When I remove containers and run again using docker-compose, MLFlow loses all the information and models I previously logged. How to make it stateful?

Here the configuration in docker-compose

  mlflow:
    image: ubuntu/mlflow:2.1.1_1.0-22.04
    command: mlflow server --host 0.0.0.0
    ports:
      - "5000:5000"
    networks:
      - mynetwork
    environment:
      - GUNICORN_CMD_ARGS=--timeout 1000
    container_name: mlflow_container
    volumes:
      - mlflow_data:/mlflow