MLFlow is an open source platform specialized in tracking ML experiments, and packaging and deploying ML models.
Current chart version is
To install the MLFlow chart in your Kubernetes cluster, add the
larribas repository (see main README), and then run:
helm install --namespace mlflow --name mlflow larribas/mlflow
After the installation succeeds, you can get the Chart’s status via:
helm status mlflow
You can delete the Chart anytime via:
helm delete --purge mlflow
I’ve created this Chart to use it in a production-ready environment in my company. We are using MLFlow with a Postgres backend store.
Therefore, the following capabilities have been left out of the Chart:
I would happily accept contributions to this Chart
By default, MLFlow will store data and artifacts in the local filesystem. If you’re deploying a production-ready MLFlow cluster, I would recommend you to point your backend store to a remote database.
At the moment, the only database engine supported by this Chart is Postgres. This means you can add the following values:
backendStore: postgres: username: my_user password: my_password host: my_host port: 5342 database: my_db
And (provided the right network and security setup) the Chart will work seamlessly with that database.
Supporting other database engine is not in my plans, but if you’re planning to fork this repository and/or contribute to it, this is what you would need to do in order to add support to another DB engine:
By default, this Chart creates a new ServiceAccount and runs the deployment under it. You can disable this behavior setting
serviceAccount.create = false.
By default, the ingress controller is disabled. You can, however, instruct the Chart to create an Ingress resource for you with the values you specify.
||Either a filepath, a database or the default value. At present, postgres is the only database engine supported by the official image. Should you want to connect to any other database, please refer to the README.|
||A local or remote filesystem path (e.g. /mnt/persistent-disk)|
||A map with the values for (username, password, host, port and database).|
||A local or remote filepath (e.g. s3://my-bucket). It is mandatory when specifying a database backend store|
||A map of arguments and values to pass to the
||The fully qualified name of the docker image to use|
||The tag for the repository (e.g. ‘latest’)|
||A list of objects. Each object should contain a
||Annotations to add to the service account|
||Specifies whether a service account should be created|
||The name of the service account to use. If not set and create is true, a name is generated using the fullname template|