metaflow-docker-deployment

Fully functional Metaflow metadata service, UI and datastore deployment with docker and docker-compose. See why Metaflow.

Image by https://metaflow.org/

Usage

First of all, make sure you have the following: python3, python3-venv, make, docker and docker-compose.

Up the containers

Run the compose file to bring up Metaflow UI, metadata/migration service, database and datastore:

$ docker-compose -f docker-compose-metaflow.yml up -d

You should be able to find that

• Metaflow UI service is available at localhost:8083
• Metaflow metadata service is available at localhost:8081
• Metaflow migration service is available at localhost:8082
• Metaflow database (postgresql) is available at localhost:10000
• Metaflow datastore (minio) console is available at localhost:10001
• Metaflow datastore (minio) S3 API is available at localhost:10002

Configure metaflow

When running flows, metaflow will look for profile configurations at ~./metaflowconfig. Let's create a new METAFLOW_PROFILE called local there, configured to use our recently deployed stack:

$ bash add_metaflow_config_local.sh

To verify that everything is working

Create a virtual environment with metaflow by running:

$ make install

Activate .venv:

$ source .venv/bin/activate

Now you can try running the hello_flow.py with the local profile. Note that we need to provide AWS access and secret keys, because metaflow will need them to connect to S3 (local minio in our case) via boto3:

$ METAFLOW_PROFILE=local AWS_ACCESS_KEY_ID=dummy_access_key AWS_SECRET_ACCESS_KEY=dummy_secret_key python hello_flow.py run

Visually inspect the results in the Metaflow UI:

Search for artifacts in the Metaflow Datastore Bucket (user: dummy_access_key, pass: dummy_secret_key):

Ping containers:

$ curl -v localhost:8081/ping localhost:8082/ping localhost:8083/ping

Terminating

To shut everything down, remember to run:

$ docker-compose -f docker-compose-metaflow.yml down