Skip to content

MLflow Installation with helm

In order to install mlflow on OpenShift using helm charts from bitnami, following steps are necessary:

  1. Open a terminal on the bastian host or direct connect to a "web terminal" client on the openshift.

  2. Create a namespace/project `mlflow in the cluster: 

    oc new-project mlflow

  3. Add the bitnami charts as a repository: 

    helm repo add bitnami https://charts.bitnami.com/bitnami

you can check if the repository is correctly added: 

helm repo list
helm rpeo update

  1. Copy the values of the mlflow charts in a yaml file: 

    helm show values bitnami/mlflow > mlflow_values.yaml

  2. Open the mlflow_values.yaml with editor (usually vim/vi): 

    vim mlflow_values.yaml

edit some parts:

The service type for tracking server is is LoadBalancer, which should be ClusterIP (Search for "LoadBalancer" in the values file you just generated): 

LoadBalancer -> ClusterIP

We will later add a route to access the UI of the mlflow tracking server.

Disable the authentification for tracking server (Search for "username: user" in the values file you just generated): 

enabled: true -> enabled: false
username: user

Due to change in bitnami policy for its images and charts, we need to pull the images from another repository.

So these repositories should be changed to find the images in bitnamilegacy instead of bitnami:

repository: bitnamil/mlflow -> bitnamilegacy/mlflow

repository: bitnamil/os-shell -> bitnamilegacy/os-shell

repository: bitnami/git -> bitnamilegacy/git

Workaround for now: In addition, we will disable the postgresql and minio:

postgresql:
  enabled: false

minio:
  enabled: false
  1. Install the MLflow using helm charts and these values: 
    helm install mlflow bitnami/mlflow --namespace mlflow --values mlflow_values.yaml