Helm Chart for mlflow

Introduction

This Helm chart installs mlflow in a Kubernetes cluster.

Prerequisites

• Kubernetes cluster 1.20+
• Helm 3+
• PV provisioner support in the underlying infrastructure.

Installation

Clone Helm chart

git clone https://github.com/zuberahmed1987/helm-mlflow.git mlflow

Configure the chart

The following items can be set via --set flag during installation or configured by editing the values.yaml directly (need to download the chart first).

Configure the way how to expose mlflow service:

• Ingress: The ingress controller must be installed in the Kubernetes cluster.
• ClusterIP: Exposes the service on a cluster-internal IP. Choosing this value makes the service only reachable from within the cluster.
• NodePort: Exposes the service on each Node’s IP at a static port (the NodePort). You’ll be able to contact the NodePort service, from outside the cluster, by requesting NodeIP:NodePort.
• LoadBalancer: Exposes the service externally using a cloud provider’s load balancer.

Install the chart

Install the mlflow helm chart with a release name my-release:

helm install --name my-release mlflow

Uninstallation

To uninstall/delete the my-release deployment:

helm delete --purge my-release

Configuration

The following table lists the configurable parameters of the mlflow chart and the default values.

Parameter	Description	Default
Image
image.repository	MLFlow Image name	ayadi05/mlflow
image.tag	MLFlow Image tag	1.5.0
image.pullPolicy	MLFlow Image pull policy	IfNotPresent
Outpout configuration
backend_store_uri	Database which store the metrics file	postgresql://user:password@postgresql.default.svc.cluster.local:5432
default_artifact_root	Bucket and path used to save the artifacts	s3://datalake/mlflow/artifacts
Database configuration
db.default.enabled	Default database (in container) enabled	false
db.host	Database host	postgresql
db.port	Database port	5432
db.user	Database user	admin
db.password	Database password	password
db.database	Database name	postgres
Minio Configuration
minio.url	mlflow S3 endpoint url frontend	minio.default.svc.cluster.local:9000
minio.accesskey	minio access key	secret
minio.secretkey	minio secret key	secret
Service
service.type	Type of service for MLFlow frontend	NodePort
service.port	Port to expose service	80
service.nodePort	Port where the service is reachable	30245
service.loadBalancerIP	LoadBalancerIP if service type is LoadBalancer	nil
service.loadBalancerSourceRanges	Address that are allowed when svc is LoadBalancer	[]
service.annotations	Service annotations	{}
Ingress
ingress.enabled	Enables Ingress	false
ingress.annotations	Ingress annotations	{}
ingress.path	Path to access frontend	/
ingress.hosts	Ingress hosts	[]
ingress.tls	Ingress TLS configuration	[]
Resources
resources	CPU/Memory resource requests/limits	{}

License

Apache License 2.0