Skip to main content

Installation and deployment

This page describes how to install the HDFS CSI driver on Kubernetes using Helm. You can use published images or build and push your own image.

Prerequisites:

  • Helm 3.x and kubectl configured
  • Access to a container registry if you publish a custom image
  • Known HDFS settings (URL, user) and Kerberos artifacts if applicable (service principal and keytab)
  1. Optional: build the Docker image

If you want to build from sources:

docker build -t <your-registry>/hdfs-csi-plugin:latest .
docker push <your-registry>/hdfs-csi-plugin:latest
  1. Prepare your Helm values

Create a custom values file (e.g., values.override.yaml):

hdfs:
  url: "hdfs://namenode:8020"
  user: "hdfs_csi"

image:
  repository: <your-registry>/hdfs-csi-plugin
  tag: latest
  1. Install

From the repository root (where the chart is located):

helm upgrade --install hdfs-csi-plugin ./hdfs-csi-plugin -n default -f values.override.yaml

Resources deployed (depending on the chart):

  • ConfigMap/Secret for HDFS and Kerberos configuration
  • Deployment/DaemonSet for Controller and Node components +- Required CSI sidecars (provisioner, attacher, registrar, liveness)
  1. Verify
kubectl get pods -n default
kubectl logs deploy/hdfs-csi-plugin -n default --tail=200
  1. Upgrade

Update your values then run:

helm upgrade hdfs-csi-plugin ./hdfs-csi-plugin -n default -f values.override.yaml
  1. Rollback
helm history hdfs-csi-plugin -n default
helm rollback hdfs-csi-plugin <REVISION> -n default
  1. Uninstall
helm uninstall hdfs-csi-plugin -n default

Deployment best practices:

  • Use a dedicated namespace (e.g., storage)
  • Set CPU/Memory requests and limits on the driver and sidecars
  • Version your values files and deploy via CI/CD
  • Validate in staging connected to a test HDFS cluster before production