Overview

The HPE CSI Driver is deployed by using industry standard means, either a Helm chart or an Operator. An "advanced install" from object configuration files is provided as reference for partners, OEMs and users wanting to perform customizations and their own packaging or deployment methodologies.

Delivery Vehicles

As different methods of installation are provided, it might not be too obvious which delivery vehicle is the right one.

Need Help Deciding?

I have a... Then you need...
Vanilla upstream Kubernetes cluster on a supported host OS. The Helm chart
Red Hat OpenShift 4.x cluster. The certified CSI operator for OpenShift
Supported environment with multiple backends. Helm chart with additional Secrets and StorageClasses
HPE Ezmeral Runtime Enterprise environment. The Helm chart
Operator Life-cycle Manager (OLM) environment. The CSI operator
Unsupported host OS/Kubernetes cluster and like to tinker. The advanced install
Supported platform in an air-gapped environment The Helm chart using the air-gapped procedure

Undecided?

If it's not clear what you should use for your environment, the Helm chart is most likely the correct answer.

Helm

Helm is the package manager for Kubernetes. Software is being delivered in a format designated as a "chart". Helm is a standalone CLI that interacts with the Kubernetes API server using your KUBECONFIG file.

The official Helm chart for the HPE CSI Driver for Kubernetes is hosted on Artifact Hub. The chart only supports Helm 3 from version 1.3.0 of the HPE CSI Driver. In an effort to avoid duplicate documentation, please see the chart for instructions on how to deploy the CSI driver using Helm.

Helm for Air-gapped Environments

In the event of deploying the HPE CSI Driver in a secure air-gapped environment, Helm is the recommended method. For sake of completeness, it's also possible to follow the advanced install procedures and replace "quay.io" in the deployment manifests with the internal private registry location.

Establish a working directory on a bastion Linux host that has HTTP access to the Internet, the private registry and the Kubernetes cluster where the CSI driver needs to be installed. The bastion host is assumed to have the docker, helm and curl command installed. It's also assumed throughout that the user executing docker has logged in to the private registry and that pulling images from the private registry is allowed anonymously by the Kubernetes compute nodes.

Note

Only the HPE CSI Driver 1.4.0 and later is supported using this methodology.

Create a working directory and set environment variables referenced throughout the procedure. In this example, we'll use HPE CSI Driver 2.3.0 on Kubernetes 1.26. Available versions are found in the co-deployments GitHub repo.

mkdir hpe-csi-driver
cd hpe-csi-driver
export MY_REGISTRY=registry.enterprise.example.com
export MY_CSI_DRIVER=2.3.0
export MY_K8S=1.26

Next, create a list with the CSI driver images. Copy and paste the entire text blob in one chunk.

curl -s https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v${MY_CSI_DRIVER}/hpe-csi-k8s-${MY_K8S}.yaml \
        https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v${MY_CSI_DRIVER}/nimble-csp.yaml \
        https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v${MY_CSI_DRIVER}/3par-primera-csp.yaml \
| grep image: | awk '{print $2}' | sort | uniq > images
echo quay.io/hpestorage/nfs-provisioner:v3.0.1 >> images

Important

The NFS server provisioner image is not automatically pulled from the private registry. Use the "nfsProvisionerImage" parameter in the StorageClass.

The above command should not output anything. A list of images should be in the file "images".

Pull, tag and push the images to the private registry.

cat images | xargs -n 1 docker pull
awk '{ print $1" "$1 }' images | sed "s/ quay.io| registry.k8s.io/ ${MY_REGISTRY}/" | xargs -n 2 docker tag
sed -e "s/quay.io|registry.k8s.io/${MY_REGISTRY}/" images | xargs -n 1 docker push

Tip

Depending on what kind of private registry being used, the base repositories hpe-storage and k8scsi might need to be created and given write access to the user pushing the images.

Next, install the chart as normal with the additional registry parameter. This is an example, please refer to the Helm chart documentation on ArtifactHub.

helm repo add hpe-storage https://hpe-storage.github.io/co-deployments/
kubectl create ns hpe-storage
helm install my-hpe-csi-driver hpe-storage/hpe-csi-driver -n hpe-storage --version ${MY_CSI_DRIVER} --set registry=${MY_REGISTRY}

Note

If the client running helm is in the air-gapped environment as well, the docs directory needs to be hosted on a web server in the air-gapped environment, and then use helm repo add hpe-storage https://my-web-server.internal/docs above instead.

Operator

The Operator pattern is based on the idea that software should be instantiated and run with a set of custom controllers in Kubernetes. It creates a native experience for any software running on Kubernetes.

The official HPE CSI Operator for Kubernetes is hosted on OperatorHub.io. The CSI Operator images are hosted both on quay.io and officially certified containers in the Red Hat Ecosystem Catalog.

Red Hat OpenShift Container Platform

The HPE CSI Operator for Kubernetes is a fully certified Operator for OpenShift. There are a few tweaks needed and there's a separate section for OpenShift.

Upstream Kubernetes and Others

Follow the documentation from the respective upstream distributions on how to deploy an Operator. In most cases, the Operator Lifecyle Manager (OLM) needs to be installed separately (does NOT apply to OpenShift 4 and later).

Visit the documentation in the OLM GitHub repo to learn how to install OLM.

Once OLM is operational, install the HPE CSI Operator.

kubectl create -f https://operatorhub.io/install/hpe-csi-operator.yaml

The Operator will be installed in my-hpe-csi-operator namespace. Watch it come up by inspecting the ClusterServiceVersion (CSV).

kubectl get csv -n my-hpe-csi-operator

Next, a HPECSIDriver object needs to be instantiated. Create a file named hpe-csi-operator.yaml and populate it according to which CSP is being deployed.

---
apiVersion: storage.hpe.com/v1
kind: HPECSIDriver
metadata:
  name: csi-driver
spec:
  disable:
    nimble: false
    primera: false
    alletra6000: false
    alletra9000: false
  imagePullPolicy: IfNotPresent
  logLevel: info
  disableNodeConformance: false
  iscsi:
    chapUser: ''
    chapPassword: ''
  registry: quay.io
  kubeletRootDir: "/var/lib/kubelet/"
  disableNodeGetVolumeStats: false
  controller:
    affinity: {}
    labels: {}
    nodeSelector: {}
    tolerations: []
  csp:
    affinity: {}
    labels: {}
    nodeSelector: {}
    tolerations: []
  node:
    affinity: {}
    labels: {}
    nodeSelector: {}
    tolerations: []

Tip

The contents depends on which version of the CSI driver being installed (the latest being illustrated). Please visit OperatorHub or ArtifactHub for more details.

Create a HPECSIDriver with the manifest.

kubectl create -f hpe-csi-operator.yaml

kubectl create -f https://scod.hpedev.io/csi_driver/examples/deployment/hpe-csi-operator.yaml

The CSI driver is now ready for use. Proceed to the next section to learn about adding an HPE storage backend.

Add an HPE Storage Backend

Once the CSI driver is deployed, two additional objects needs to be created to get started with dynamic provisioning of persistent storage, a Secret and a StorageClass.

Tip

Naming the Secret and StorageClass is entirely up to the user, however, to keep up with the examples on SCOD, it's highly recommended to use the names illustrated here.

Secret Parameters

All parameters are mandatory and described below.

Parameter Description
serviceName This hostname or IP address where the Container Storage Provider (CSP) is running, usually a Kubernetes Service, such as "alletra6000-csp-svc" or "alletra9000-csp-svc"
servicePort This is port the serviceName is listening to.
backend This is the management hostname or IP address of the actual backend storage system, such as an Alletra 5000/6000 or 9000 array.
username Backend storage system username with the correct privileges to perform storage management.
password Backend storage system password.

Example:

apiVersion: v1
kind: Secret
metadata:
  name: hpe-backend
  namespace: hpe-storage
stringData:
  serviceName: alletra6000-csp-svc
  servicePort: "8080"
  backend: 192.168.1.110
  username: admin
  password: admin

apiVersion: v1
kind: Secret
metadata:
  name: hpe-backend
  namespace: hpe-storage
stringData:
  serviceName: alletra9000-csp-svc
  servicePort: "8080"
  backend: 10.10.0.20
  username: 3paradm
  password: 3pardata

apiVersion: v1
kind: Secret
metadata:
  name: hpe-backend
  namespace: hpe-storage
stringData:
  serviceName: nimble-csp-svc
  servicePort: "8080"
  backend: 192.168.1.2
  username: admin
  password: admin

apiVersion: v1
kind: Secret
metadata:
  name: hpe-backend
  namespace: hpe-storage
stringData:
  serviceName: primera3par-csp-svc
  servicePort: "8080"
  backend: 10.10.0.2
  username: 3paradm
  password: 3pardata

Create the Secret using kubectl:

kubectl create -f secret.yaml

Tip

In a real world scenario it's more practical to name the Secret something that makes sense for the organization. It could be the hostname of the backend or the role it carries, i.e "hpe-alletra-sanjose-prod".

Next step involves creating a default StorageClass.

Configuring Additional Storage Backends

It's not uncommon to have multiple HPE primary storage systems within the same environment, either the same family or different ones. This section walks through the scenario of managing multiple StorageClass and Secret API objects to represent an environment with multiple systems.

There's a brief tutorial available in the Video Gallery that walks through these steps.

Note

Make note of the Kubernetes Namespace or OpenShift project name used during the deployment. In the following examples, we will be using the "hpe-storage" Namespace.

To view the current Secrets in the "hpe-storage" Namespace (assuming default names):

kubectl -n hpe-storage get secret/hpe-backend
NAME                     TYPE          DATA      AGE
hpe-backend              Opaque        5         2m

This Secret is used by the CSI sidecars in the StorageClass to authenticate to a specific backend for CSI operations. In order to add a new Secret or manage access to multiple backends, additional Secrets will need to be created per backend.

Secret Requirements

  • Each Secret name must be unique.
  • servicePort should be set to 8080.

To create a new Secret, specify the name, Namespace, backend username, backend password and the backend IP address to be used by the CSP and save it as custom-secret.yaml (a detailed description of the parameters are available above).

apiVersion: v1
kind: Secret
metadata:
  name: custom-secret
  namespace: hpe-storage
stringData:
  serviceName: alletra6000-csp-svc
  servicePort: "8080"
  backend: 192.168.1.110
  username: admin
  password: admin

apiVersion: v1
kind: Secret
metadata:
  name: custom-secret
  namespace: hpe-storage
stringData:
  serviceName: alletra9000-csp-svc
  servicePort: "8080"
  backend: 10.10.0.20
  username: 3paradm
  password: 3pardata

apiVersion: v1
kind: Secret
metadata:
  name: custom-secret
  namespace: hpe-storage
stringData:
  serviceName: nimble-csp-svc
  servicePort: "8080"
  backend: 192.168.1.2
  username: admin
  password: admin

apiVersion: v1
kind: Secret
metadata:
  name: custom-secret
  namespace: hpe-storage
stringData:
  serviceName: primera3par-csp-svc
  servicePort: "8080"
  backend: 10.10.0.2
  username: 3paradm
  password: 3pardata

Create the Secret using kubectl:

kubectl create -f custom-secret.yaml

You should now see the Secret in the "hpe-storage" Namespace:

kubectl -n hpe-storage get secret/custom-secret
NAME                     TYPE          DATA      AGE
custom-secret            Opaque        5         1m

Create a StorageClass with the Custom Secret

To use the new Secret "custom-secret", create a new StorageClass using the Secret and the necessary StorageClass parameters. Please see the requirements section of the respective CSP.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: hpe-custom
provisioner: csi.hpe.com
parameters:
  csi.storage.k8s.io/fstype: xfs
  csi.storage.k8s.io/controller-expand-secret-name: custom-secret
  csi.storage.k8s.io/controller-expand-secret-namespace: hpe-storage
  csi.storage.k8s.io/controller-publish-secret-name: custom-secret
  csi.storage.k8s.io/controller-publish-secret-namespace: hpe-storage
  csi.storage.k8s.io/node-publish-secret-name: custom-secret
  csi.storage.k8s.io/node-publish-secret-namespace: hpe-storage
  csi.storage.k8s.io/node-stage-secret-name: custom-secret
  csi.storage.k8s.io/node-stage-secret-namespace: hpe-storage
  csi.storage.k8s.io/provisioner-secret-name: custom-secret
  csi.storage.k8s.io/provisioner-secret-namespace: hpe-storage
  description: "Volume created by using a custom Secret with the HPE CSI Driver for Kubernetes"
reclaimPolicy: Delete
allowVolumeExpansion: true

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: hpe-custom
provisioner: csi.hpe.com
parameters:
  csi.storage.k8s.io/fstype: xfs
  csi.storage.k8s.io/resizer-secret-name: custom-secret
  csi.storage.k8s.io/resizer-secret-namespace: hpe-storage
  csi.storage.k8s.io/controller-publish-secret-name: custom-secret
  csi.storage.k8s.io/controller-publish-secret-namespace: hpe-storage
  csi.storage.k8s.io/node-publish-secret-name: custom-secret
  csi.storage.k8s.io/node-publish-secret-namespace: hpe-storage
  csi.storage.k8s.io/node-stage-secret-name: custom-secret
  csi.storage.k8s.io/node-stage-secret-namespace: hpe-storage
  csi.storage.k8s.io/provisioner-secret-name: custom-secret
  csi.storage.k8s.io/provisioner-secret-namespace: hpe-storage
  description: "Volume created by using a custom Secret with the HPE CSI Driver for Kubernetes"
reclaimPolicy: Delete
allowVolumeExpansion: true

Note

Don't forget to call out the StorageClass explicitly when creating PVCs from non-default StorageClasses.

Next, Create a PersistentVolumeClaim from a StorageClass.

Advanced Install

This guide is primarily written to accommodate a highly manual installation on upstream Kubernetes or partner OEMs engaged with HPE to bundle the HPE CSI Driver in a custom distribution. Installation steps may vary for different vendors and flavors of Kubernetes.

The following example walks through deployment of the latest CSI driver.

Critical

It's highly recommended to use either the Helm chart or Operator to install the HPE CSI Driver for Kubernetes and the associated Container Storage Providers. Only venture down manual installation if your requirements can't be met by the Helm chart or Operator.

Manual CSI Driver Install

Deploy the CSI driver and sidecars for the relevant Kubernetes version.

Uninstalling the CSI driver when installed manually

The manifests below create a number of objects, including CustomResourceDefinitions (CRDs) which may hold critical information about storage resources. Simply deleting the below manifests in order to uninstall the CSI driver may render PersistentVolumes unusable.

Common

These object configuration files are common for all versions of Kubernetes.

All components below are deployed in the "hpe-storage" Namespace.

kubectl create ns hpe-storage

Worker node IO settings:

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-linux-config.yaml

Container Storage Provider:

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/nimble-csp.yaml

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/3par-primera-csp.yaml
kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/3par-primera-crd.yaml

Important

The above instructions assumes you have an array with a supported platform OS installed. Please see the requirements section of the respective CSP.

Install the CSI driver:

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.26.yaml

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.25.yaml

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.24.yaml

kubectl apply -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.23.yaml

Seealso

Older and unsupported versions of Kubernetes and the CSI driver are archived on this page.

Depending on which version is being deployed, different API objects gets created. Next step: Add an HPE Storage Backend.

Advanced Uninstall

The following steps outline how to uninstall the CSI driver that has been deployed using the Advanced Install above.

Uninstall Worker node settings:

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-linux-config.yaml

Uninstall relevant Container Storage Provider:

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/nimble-csp.yaml

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/3par-primera-csp.yaml

HPE Alletra 9000, Primera and 3PAR users

If you are reinstalling the HPE CSI Driver, DO NOT remove the crd/hpevolumeinfos.storage.hpe.com resource. This CustomResourceDefinition contains important volume metadata used by the HPE Alletra 9000, Primera and 3PAR CSP. HPE CSI Driver v2.0.0 and below share the same YAML file for crds and CSP and would require a manual removal of the individual Service and Deployment in the "hpe-storage" Namespace.

Uninstall the CSI driver:

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.26.yaml

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.26.yaml

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.24.yaml

kubectl delete -f https://raw.githubusercontent.com/hpe-storage/co-deployments/master/yaml/csi-driver/v2.3.0/hpe-csi-k8s-1.23.yaml

If no longer needed, delete the "hpe-storage" Namespace.

kubectl delete ns hpe-storage

Downgrading the CSI driver

Downgrading the CSI driver is currently not supported. It will work between certain minor versions. HPE does not test or document procedures to downgrade between incompatible versions.