OpenStack Cinder CSI volume provisioner

This guide is designed to help you effortlessly integrate the Cinder CSI Volume Provisioner into your OKD or OpenShift cluster.

Niklas Hagman

Niklas Hagman

OpenShift Platform Specialist

Have you configured your OKD or OpenShift cluster with the platform option set to "none" and, as a result, are missing the OpenStack Cinder CSI Driver Operator?

This guide is designed to help you effortlessly integrate the Cinder CSI Volume Provisioner into your OKD or OpenShift cluster. It’s designed to streamline the process, making the integration seamless and hassle-free.

Moreover, this guide is not exclusive to OpenShift or OKD environments; it can be easily adapted for use in vanilla Kubernetes setups with a simple modification. A key highlight for OKD and OpenShift users is the inclusion of Security Context Constraints (SCC) cluster role bindings in the Helm chart’s templates directory. This critical feature enables the Cinder CSI pods to run with privileged access, aligning them with OpenShift’s security practices and ensuring their optimal functionality within your cluster’s security framework.

Throughout this guide, we will take you step by step through the installation process, providing clear instructions and helpful tips to ensure a successful integration. Whether you’re new to OpenShift or an experienced administrator, this guide aims to provide you with all the necessary information to enhance your cluster’s storage capabilities with the Cinder CSI Volume Provisioner.

Git-repository

This guide is accompanied by a complete set of code examples available on our GitHub repository. For easy access to all the scripts, configurations, and templates used throughout this guide, please visit safespring-community/utilities on GitHub. This will ensure you have everything you need to successfully integrate the Cinder CSI Volume Provisioner into your environment.

Setting up your secret for OpenStack authentication

Securing communication between Cinder CSI volume provisioner and OpenStack is paramount. Utilizing an application credential facilitates this by providing the necessary authentication details for interactions with OpenStack services.

1. Creating an application credential

Initiate the creation of a new application credential tailored for your requirements. If you have previously generated credentials, consider reusing them for the Cinder CSI plugin. To manage your credentials effectively, use the command below to create a new set or list existing ones:

openstack application credential create <app-cred-name>
openstack application credential list

Upon creation, extract the auth_url, Application ID, and Application secret to enable OpenStack authentication:

auth_url=$(openstack configuration show -f json | jq .auth_url)

json_output=$(openstack application credential create cinder-csi --format json)
app_id=$(echo $json_output | jq -r '.id')
app_secret=$(echo $json_output | jq -r '.secret')

echo "auth_url: ${auth_url}"
echo "Application ID: ${app_id}"
echo "Application secret: ${app_secret}"

2. Create a namespace for storage CSI

Establish a dedicated namespace for the CSI, enhancing organizational clarity and security:

namespace="csi"
oc create namespace ${namespace}

3. Prepare Your cloud configuration

Generate a configuration file encoded in base64 for Kubernetes secret storage. This configuration allows your application to authenticate with OpenStack:

cloud_config="[Global]
auth-url=${auth_url}
application-credential-id=${app_id}
application-credential-secret=${app_secret}"

cloud_config_encoded=$(echo "${cloud_config}" | base64 | tr -d '\n')
echo -e "${cloud_config_encoded}" | base64 -d

4. Deploy the encoded configuration as your secret

Utilize this Kubernetes manifest to securely store your OpenStack credentials within the created namespace:

oc apply -f - <<EOF
kind: Secret
apiVersion: v1
metadata:
  name: cinder-csi-cloud-config
  namespace: ${namespace}
data:
  cloud.conf: ${cloud_config_encoded}
type: Opaque
EOF

Installation process

Selecting the correct Cinder CSI volume provisioner version

Match the Cinder CSI provisioner version with your Kubernetes version. Retrieve your Kubernetes version and list available Helm chart versions to ensure compatibility.

Obtain your OpenShift version:

oc version
Client Version: 4.15.0
Kustomize Version: v5.0.4-0.20230601165947-6ce0bf390ce3
Server Version: 4.15.0
Kubernetes Version: v1.28.6+6216ea1

List available Cinder CSI versions:

namespace="${namespace:-csi}"
helm repo -n ${namespace} add cpo https://kubernetes.github.io/cloud-provider-openstack
helm search -n ${namespace} repo cpo/openstack-cinder-csi --versions
name version app_version description
cpo/openstack-cinder-csi 2.29.0 v1.29.0 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.28.2 v1.28.2 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.28.1 v1.28.1 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.28.0 v1.28.0 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.27.3 v1.27.3 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.27.2 v1.27.2 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.27.1 v1.27.1 Cinder CSI Chart for OpenStack
cpo/openstack-cinder-csi 2.27.0 v1.27.0 Cinder CSI Chart for OpenStack

Match your kubernetes version with the version in app_version column. When searching for specific version, keep in mind that it is the Helm chart version you are specifying, not the kubernetes version.

helm search -n ${namespace} repo cpo/openstack-cinder-csi --version '~2.28'
name version app_version description
cpo/openstack-cinder-csi 2.28.2 v1.28.2 Cinder CSI Chart for OpenStack

When searching for the right version, we are using Tilde range comparisons. The tilde (~) comparison operator is for patch level ranges when a minor version is specified and major level changes when the minor number is missing. In our example, ~2.28 is equivalent to >= 2.28, < 2.29.

In Chart.yaml, update the dependencies so it searching for the same version

dependencies:
  - name: openstack-cinder-csi
    version: '~2.28'
    repository: "https://kubernetes.github.io/cloud-provider-openstack"

Update dependencies

helm dependency update

If you have changes, push this to your git repository.

Installing Cinder CSI volume provisioner

namespace="${namespace:-csi}"
helm install -n ${namespace} cinder-csi .

Verify with oc -n ${namespace} get pods that you have one controllerplugin pod and nodeplugin pods for each node in your cluster.

oc -n ${namespace} get pods
NAME READY STATUS RESTARTS AGE
openstack-cinder-csi-controllerplugin-544fc6fc4c-cnjft 6/6 Running 0 151m
openstack-cinder-csi-nodeplugin-5t54r 3/3 Running 0 151m
openstack-cinder-csi-nodeplugin-dc5hc 3/3 Running 0 151m
openstack-cinder-csi-nodeplugin-dxkhb 3/3 Running 0 151m
openstack-cinder-csi-nodeplugin-kxzr8 3/3 Running 0 151m
openstack-cinder-csi-nodeplugin-vp8qg 3/3 Running 0 151m

You now have two different storage classes to use.

oc get storageclass -o custom-columns=Name:.metadata.name,Provisoner:.provisioner
Name Provisoner
csi-cinder-sc-delete cinder.csi.openstack.org
csi-cinder-sc-retain cinder.csi.openstack.org

Test Cinder CSI volume provisioner

Test Cinder CSI volume provisioner by creating a Persistent Volume Claim and then a application that’s using this PVC.

namespace_test="csi-test"
oc create namespace ${namespace_test}
oc apply -f - <<EOF
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: csi-pvc-cinderplugin
  namespace: ${namespace_test}
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  storageClassName: csi-cinder-sc-delete
EOF
oc apply -f - <<EOF
apiVersion: v1
kind: Pod
metadata:
  name: nginx
  namespace: ${namespace_test}
spec:
  containers:
    - image: docker.io/nginxinc/nginx-unprivileged
      imagePullPolicy: IfNotPresent
      name: nginx
      ports:
        - containerPort: 8080
          protocol: TCP
      volumeMounts:
        - mountPath: /var/lib/www/html
          name: csi-data-cinderplugin
  volumes:
    - name: csi-data-cinderplugin
      persistentVolumeClaim:
        claimName: csi-pvc-cinderplugin
        readOnly: false
EOF

The persistent volume claim should be “Bound” when successful.

oc -n ${namespace_test} get pvc -o custom-columns=Name:.metadata.name,Status:.status.phase,Volume:.spec.volumeName
Name Status Volume
csi-pvc-cinderplugin Bound pvc-ed60e725-93e8-447c-bc18-ca33546f2ce8

If you have any problems, start by looking into the events table with oc -n ${namespace_test} events.

In Openstack you can use openstack cli to see your volume.

openstack volume show pvc-ed60e725-93e8-447c-bc18-ca33546f2ce8

Finally delete your test:

oc delete namespace ${namespace_test}

Updating your installation

Should there be updates available for the Cinder CSI provisioner, use the following command to apply them:

namespace="${namespace:-csi}"
helm upgrade -n ${namespace} cinder-csi .

Uninstalling

If you need to uninstall the Cinder CSI provisioner, execute these commands:

namespace="${namespace:-csi}"
helm install -n ${namespace} cinder-csi .
oc delete namespace ${namespace}

FAQ

The challenge with using Helm’s values.yaml for secrets

While Helm charts offer the convenience of automating deployments, including the creation of secrets via the values.yaml file, this method presents significant security challenges, especially in a GitOps workflow with ArgoCD.

For example, when configuring the openstack-cinder-csi chart, you might be tempted to directly embed sensitive credentials within the values.yaml file like so:

openstack-cinder-csi:
  secret:
    enabled: true
    create: true
    name: cinder-csi-cloud-config
    data:
      cloud.conf: |-
        [Global]
        auth-url=<auth_url>
        application-credential-id=<app_id>
        application-credential-secret=<app_secret>        

This approach works fine for manual Helm installations. However, in a GitOps setup where ArgoCD automatically applies configurations from a Git repository, storing secrets in this manner is risky. The primary concern is security: storing sensitive data, like credentials, in a Git repository—even if it’s private—exposes your infrastructure to potential breaches.