MLflow Production Docker

A production ready¹ docker-compose deployment of MLflow using Traefik with HTTP Basic Auth.

¹ This project is provided as is, and without a warranty of any kind, see LICENSE for full details.

Setup

Install requirements:

• Apache Utils (provides htpasswd command)
• GNU gettext (provides envsubst command)

sudo apt-get install -y apache2-utils gettext

Create a new .htpasswd file with a single user (using Bcrypt hashing):

htpasswd -cB config/.htpasswd <username>

Add additional users with:

htpasswd -B config/.htpasswd <other_username>

Create a .env in the root directory with the following variables completed:

# connection settings
DOMAIN_NAME=localhost
SERVICE_NAME=mlflow
SERVICE_PORT=5000
# tell the user where they are logging into
REALM=testsite
# required for letsencrypt certificate email updates
EMAIL_ADDRESS=example@example.com

Finally write a traefik.yml config file using the .env variables:

./write_template.sh

Running

Start all containers:

docker-compose up -d --build

MLflow is accessible on port 443 over HTTPS (port 80 redirects to 443). Use the username and password you created above to log in.