CSI Driver installation using Dell CSI Operator

Installation of CSI drivers using Dell CSI Operator

The Dell CSI Operator is no longer actively maintained or supported. Dell CSI Operator has been replaced with Dell CSM Operator. If you are currently using Dell CSI Operator, refer to the operator migration documentation to migrate from Dell CSI Operator to Dell CSM Operator.

The Dell CSI Operator is a Kubernetes Operator, which can be used to install and manage the CSI Drivers provided by Dell for various storage platforms. This operator is available as a community operator for upstream Kubernetes and can be deployed using OperatorHub.io. It is also available as a certified operator for OpenShift clusters and can be deployed using the OpenShift Container Platform. Both these methods of installation use OLM (Operator Lifecycle Manager). The operator can also be deployed manually.


(Optional) Volume Snapshot Requirements

On Upstream Kubernetes clusters, ensure that to install

  • VolumeSnapshot CRDs - Install v1 VolumeSnapshot CRDs
  • External Volume Snapshot Controller

For detailed snapshot setup procedure, click here.

NOTE: That step can be skipped with OpenShift.


Dell CSI Operator has been tested and qualified with

Before you begin

If you have installed an old version of the dell-csi-operator which was available with the name CSI Operator, please refer to this section before continuing.

Full list of CSI Drivers and versions supported by the Dell CSI Operator

CSI Driver Version ConfigVersion Kubernetes Version OpenShift Version
CSI PowerMax 2.5.0 v2.5.0 1.23, 1.24, 1.25 4.10, 4.10 EUS, 4.11
CSI PowerMax 2.6.0 v2.6.0 1.24, 1.25, 1.26 4.10, 4.10 EUS, 4.11
CSI PowerMax 2.7.0 v2.7.0 1.25, 1.26, 1.27 4.11, 4.12, 4.12 EUS
CSI PowerFlex 2.5.0 v2.5.0 1.23, 1.24, 1.25 4.10, 4.10 EUS, 4.11
CSI PowerFlex 2.6.0 v2.6.0 1.24, 1.25, 1.26 4.10, 4.10 EUS, 4.11
CSI PowerFlex 2.7.0 v2.7.0 1.25, 1.26, 1.27 4.11, 4.12 EUS, 4.12
CSI PowerScale 2.5.0 v2.5.0 1.23, 1.24, 1.25 4.10, 4.10 EUS, 4.11
CSI PowerScale 2.6.0 v2.6.0 1.24, 1.25, 1.26 4.10, 4.10 EUS, 4.11
CSI PowerScale 2.7.0 v2.7.0 1.25, 1.26, 1.27 4.11, 4.12, 4.12 EUS
CSI Unity XT 2.5.0 v2.5.0 1.23, 1.24, 1.25 4.10, 4.10 EUS, 4.11
CSI Unity XT 2.6.0 v2.6.0 1.24, 1.25, 1.26 4.10, 4.10 EUS, 4.11
CSI Unity XT 2.7.0 v2.7.0 1.25, 1.26, 1.27 4.11, 4.12, 4.12 EUS
CSI PowerStore 2.5.0 v2.5.0 1.23, 1.24, 1.25 4.10, 4.10 EUS, 4.11
CSI PowerStore 2.6.0 v2.6.0 1.24, 1.25, 1.26 4.10, 4.10 EUS, 4.11
CSI PowerStore 2.7.0 v2.7.0 1.25, 1.26, 1.27 4.11, 4.12, 4.12 EUS

Dell CSI Operator can be installed via OLM (Operator Lifecycle Manager) and manual installation.

Installation Using Operator Lifecycle Manager

dell-csi-operator can be installed using Operator Lifecycle Manager (OLM) on upstream Kubernetes clusters & Red Hat OpenShift Clusters.
The installation process involves the creation of a Subscription object either via the OperatorHub UI or using kubectl/oc. While creating the Subscription you can set the Approval strategy for the InstallPlan for the Operator to -

  • Automatic - If you want the Operator to be automatically installed or upgraded (once an upgrade becomes available)
  • Manual - If you want a Cluster Administrator to manually review and approve the InstallPlan for installation/upgrades

NOTE: The recommended version of OLM for upstream Kubernetes is v0.18.3.

Pre-Requisite for installation with OLM

Please run the following commands for creating the required ConfigMap before installing the dell-csi-operator using OLM.
#Replace operator-namespace in the below command with the actual namespace where the operator will be deployed by OLM

git clone -b v1.12.0 https://github.com/dell/dell-csi-operator.git
cd dell-csi-operator
tar -czf config.tar.gz driverconfig/
kubectl create configmap dell-csi-operator-config --from-file config.tar.gz -n <operator-namespace>
Upstream Kubernetes
Red Hat OpenShift Clusters
  • For installing via OpenShift with the Operator, go to the OpenShift page.

Manual Installation


Skip step 1 for “offline bundle installation” and continue using the workspace created by untar of dell-csi-operator-bundle.tar.gz.

  1. Clone and checkout the required dell-csi-operator version using git clone -b v1.12.0 https://github.com/dell/dell-csi-operator.git.
  2. cd dell-csi-operator
  3. Run bash scripts/install.sh to install the operator.
  1. Run the command oc get pods -n dell-csi-operator to validate the installation. If completed successfully, you should be able to see the operator-related pod in the ‘dell-csi-operator’ namespace.

Custom Resource Definitions

As part of the Dell CSI Operator installation, a CRD representing each driver installation is also installed.
List of CRDs which are installed in API Group storage.dell.com

  • csipowermax
  • csiunity
  • csivxflexos
  • csiisilon
  • csipowerstore
  • csipowermaxrevproxy

For installation of the supported drivers, a CustomResource has to be created in your cluster.

Pre-Requisites for installation of the CSI Drivers

Pre-requisites for upstream Kubernetes Clusters

On upstream Kubernetes clusters, make sure to install

  • VolumeSnapshot CRDs
    • On clusters running v1.25,v1.26 & v1.27, make sure to install v1 VolumeSnapshot CRDs
  • External Volume Snapshot Controller with the correct version

Pre-requisites for Red Hat OpenShift Clusters


If you are installing a CSI driver which is going to use iSCSI as the transport protocol, please follow the following instructions.
In Red Hat OpenShift clusters, you can create a MachineConfig object using the console or oc to ensure that the iSCSI daemon starts on all the Red Hat CoreOS nodes. Here is an example of a MachineConfig object:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
  name: 99-iscsid
    machineconfiguration.openshift.io/role: worker
      version: 3.2.0  
      - name: "iscsid.service"
        enabled: true

Once the MachineConfig object has been deployed, CoreOS will ensure that iscsid.service starts automatically.

Alternatively, you can check the status of the iSCSI service by entering the following command on each worker node in the cluster:

sudo systemctl status iscsid

The service should be up and running (i.e. should be active state).

If the iscsid.service is not running, then perform the following steps on each worker node in the cluster

  1. Login to worker nodes and check if the file /etc/iscsi/initiatorname.iscsi has been created properly
  2. If the file doesn’t exist or it doesn’t contain a valid ISCSI IQN, then make sure it exists with valid entries
  3. Ensure that iscsid service is running - Enable sudo systemctl enable iscsid & restart sudo systemctl restart iscsid iscsid if necessary. Note: If your worker nodes are running Red Hat CoreOS, make sure that automatic ISCSI login at boot is configured. Please contact RedHat for more details.


If you are installing a CSI Driver which requires the installation of the Linux native Multipath software - multipathd, please follow the below instructions

To enable multipathd on RedHat CoreOS nodes you need to prepare a working configuration encoded in base64.

user_friendly_names yes
find_multipaths yes
blacklist {
}' | base64 -w0

Use the base64 encoded string output in the following MachineConfig yaml file (under source section)

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
  name: workers-multipath-conf-default
    machineconfiguration.openshift.io/role: worker
      version: 3.2.0
      - contents:
          source: data:text/plain;charset=utf-8;base64,ZGVmYXVsdHMgewp1c2VyX2ZyaWVuZGx5X25hbWVzIHllcwpmaW5kX211bHRpcGF0aHMgeWVzCn0KCmJsYWNrbGlzdCB7Cn0K
          verification: {}
        filesystem: root
        mode: 400
        path: /etc/multipath.conf

After deploying thisMachineConfig object, CoreOS will start multipath service automatically.
Alternatively, you can check the status of the multipath service by entering the following command in each worker nodes.
sudo multipath -ll

If the above command is not successful, ensure that the /etc/multipath.conf file is present and configured properly. Once the file has been configured correctly, enable the multipath service by running the following command: sudo /sbin/mpathconf –-enable --with_multipathd y

Finally, you have to restart the service by providing the command sudo systemctl restart multipathd

For additional information refer to official documentation of the multipath configuration.

Installing CSI Driver via Operator

CSI Drivers can be installed by creating a CustomResource object in your cluster.

Sample manifest files for each driver CustomResourceDefintion have been provided in the samples folder to help with the installation of the drivers. These files follow the naming convention

  {driver name}_{driver version}_k8s_{k8 version}.yaml


  {driver name}_{driver version}_ops_{OpenShift version}.yaml

For e.g.

  • samples/powermax_v270_k8s_127.yaml* <- To install CSI PowerMax driver v2.7.0 on a Kubernetes 1.27 cluster
  • samples/powermax_v270_ops_412.yaml* <- To install CSI PowerMax driver v2.7.0 on an OpenShift 4.12 cluster

Copy the correct sample file and edit the mandatory & any optional parameters specific to your driver installation by following the instructions here

NOTE: A detailed explanation of the various mandatory and optional fields in the CustomResource is available here. Please make sure to read through and understand the various fields.

Run the following command to install the CSI driver.

kubectl create -f <driver-manifest.yaml>

Note: If you are using an OLM based installation, the example manifests are available in the OperatorHub UI. You can edit these manifests and install the driver using the OperatorHub UI.

Verifying the installation

Once the driver Custom Resource has been created, you can verify the installation

  • Check if Driver CR got created successfully

    For e.g. – If you installed the PowerMax driver

    kubectl get csipowermax -n <driver-namespace>
  • Check the status of the Custom Resource to verify if the driver installation was successful

If the driver-namespace was set to test-powermax, and the name of the driver is powermax, then run the command kubectl get csipowermax/powermax -n test-powermax -o yaml to get the details of the Custom Resource.

Note: If the state of the CustomResource is Running then all the driver pods have been successfully installed. If the state is SuccessFul, then it means the driver deployment was successful but some driver pods may not be in a Running state. Please refer to the Troubleshooting section here if you encounter any issues during installation.

Update CSI Drivers

The CSI Drivers installed by the Dell CSI Operator can be updated like any Kubernetes resource. This can be achieved in various ways which include –

  • Modifying the installation directly via kubectl edit

    kubectl get <driver-object> -n <driver-namespace>

    For example - If the Unity XT driver is installed then run this command to get the object name of kind CSIUnity. #Replace driver-namespace with the namespace where the Unity XT driver is installed

    kubectl get csiunity -n <driver-namespace>

    use the object name in kubectl edit command.

    kubectl edit <driver-object>/<object-name> -n <driver-namespace>

    For example - If the object name is CSIUnity. #Replace object-name with the object name of kind CSIUnity

    kubectl edit csiunity/<object-name> -n <driver-namespace>

    and modify the installation. The usual fields to edit are the version of drivers, sidecars and the environment variables.

  • Modify the API object in place via kubectl patch command. For example if you want to patch the deployment to have two replicas for Unity XT driver then run this command to get the deployment

    kubectl get deployments -n <driver-namespace>

    to patch the deployment with your patch object inline run this command. #Replace deployment with the name of the deployment

    kubectl patch deploy/<deployment> -n <driver-namespace> -p '{"spec":{"replicas": 2}}'

    to patch the deployment with your patch file run this command. #Replace deployment with the name of the deployment

    kubectl patch deployment <deployment> --patch-file patch-file.yaml

To create patch file or edit deployments, refer here for driver version & environment variables and here for version of side-cars. The latest versions of drivers could have additional environment variables or sidecars.

The below notes explain some of the general items to take care of.


  1. If you are trying to upgrade the CSI driver from an older version, make sure to modify the configVersion field if required.
         configVersion: v2.7.0
  2. Volume Health Monitoring feature is optional and by default this feature is disabled for drivers when installed via operator. To enable this feature, we will have to modify the below block while upgrading the driver.To get the volume health state add external-health-monitor sidecar in the sidecar section and valueunder controller set to true and the value under node set to true as shown below:
    i. Add controller and node section as below:
            - name: X_CSI_HEALTH_MONITOR_ENABLED
              value: "true"
        dnsPolicy: ClusterFirstWithHostNet
            - name: X_CSI_HEALTH_MONITOR_ENABLED
              value: "true"
    ii. Update the sidecar versions and add external-health-monitor sidecar if you want to enable health monitor of CSI volumes from Controller plugin:
        - args:
          - --volume-name-prefix=csiunity
          - --default-fstype=ext4
          image: k8s.gcr.io/sig-storage/csi-provisioner:v3.4.0
          imagePullPolicy: IfNotPresent
          name: provisioner
        - args:
          - --snapshot-name-prefix=csiunitysnap
          image: k8s.gcr.io/sig-storage/csi-snapshotter:v6.2.1
          imagePullPolicy: IfNotPresent
          name: snapshotter
        - args:
          - --monitor-interval=60s
          image: gcr.io/k8s-staging-sig-storage/csi-external-health-monitor-controller:v0.8.0
          imagePullPolicy: IfNotPresent
          name: external-health-monitor
        - image: k8s.gcr.io/sig-storage/csi-attacher:v4.2.0
          imagePullPolicy: IfNotPresent
          name: attacher
        - image: k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.6.3
          imagePullPolicy: IfNotPresent
          name: registrar
        - image: k8s.gcr.io/sig-storage/csi-resizer:v1.7.0
          imagePullPolicy: IfNotPresent
          name: resizer
  3. Configmap needs to be created with command kubectl create -f configmap.yaml using following yaml file.
kind: ConfigMap
  name: unity-config-params
  namespace: unity
  driver-config-params.yaml: |
    CSI_LOG_LEVEL: "info"
    TENANT_NAME: ""    

NOTE: Replicas in the driver CR file should not be greater than or equal to the number of worker nodes when upgrading the driver. If the Replicas count is not less than the worker node count, some of the driver controller pods would land in a pending state, and upgrade will not be successful. Driver controller pods go in a pending state because they have anti-affinity to each other and cannot be scheduled on nodes where there is a driver controller pod already running. Refer to https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/#inter-pod-affinity-and-anti-affinity for more details.

NOTE: Do not try to update the operator by modifying the original CustomResource manifest file and running the kubectl apply -f command. As part of the driver installation, the Operator sets some annotations on the CustomResource object which are further utilized in some workflows (like detecting upgrade of drivers). If you run the kubectl apply -f command to update the driver, these annotations are overwritten and this may lead to failures.

NOTE: From v1.4.0 onwards, Dell CSI Operator does not support the creation of StorageClass and VolumeSnapshotClass objects. Although these fields are still present in the various driver CustomResourceDefinitions, they would be ignored by the operator. These fields will be removed from the CustomResourceDefinitions in a future release. If StorageClass and VolumeSnapshotClass need to be retained, you should upgrade the driver as per the recommended way noted above. StorageClass and VolumeSnapshotClass would not be retained on driver uninstallation.

Supported modifications

  • Changing environment variable values for driver
  • Adding (supported) environment variables
  • Updating the image of the driver


  • The Dell CSI Operator can’t manage any existing driver installed using Helm charts. If you already have installed one of the Dell CSI drivers in your cluster and want to use the operator based deployment, uninstall the driver and then redeploy the driver following the installation procedure described.
  • The Dell CSI Operator is not fully compliant with the OperatorHub React UI elements and some of the Custom Resource fields may show up as invalid or unsupported in the OperatorHub GUI. To get around this problem, use kubectl/oc commands to get details about the Custom Resource(CR). This issue will be fixed in the upcoming releases of the Dell CSI Operator

Custom Resource Specification

Each CSI Driver installation is represented by a Custom Resource.

The specification for the Custom Resource is the same for all the drivers.
Below is a list of all the mandatory and optional fields in the Custom Resource specification

Mandatory fields

configVersion - Configuration version - Refer full list of supported driver for finding out the appropriate config version here
replicas - Number of replicas for controller plugin - Must be set to 1 for all drivers
dnsPolicy - Determines the dnsPolicy for the node daemonset. Accepted values are Default, ClusterFirst, ClusterFirstWithHostNet, None common
This field is mandatory and is used to specify common properties for both controller and the node plugin

  • image - driver container image
  • imagePullPolicy - Image Pull Policy of the driver image
  • envs - List of environment variables and their values

Optional fields

controller - List of environment variables and values which are applicable only for controller
node - List of environment variables and values which are applicable only for node
sideCars - Specification for CSI sidecar containers.
authSecret - Name of the secret holding credentials for use by the driver. If not specified, the default secret *-creds must exist in the same namespace as driver
tlsCertSecret - Name of the TLS cert secret for use by the driver. If not specified, a secret *-certs must exist in the namespace as driver

Boolean value which can be set to true in order to force update the status of the CSI Driver

tolerations List of tolerations which should be applied to the driver StatefulSet/Deployment and DaemonSet
It should be set separately in the controller and node sections if you want separate set of tolerations for them

nodeSelector Used to specify node selectors for the driver StatefulSet/Deployment and DaemonSet

fsGroupPolicy Defines which FS Group policy mode to be used, Supported modes: None, File and ReadWriteOnceWithFSType

Here is a sample specification annotated with comments to explain each field

apiVersion: storage.dell.com/v1
kind: CSIPowerMax          # Type of the driver
  name: test-powermax      # Name of the driver
  namespace: test-powermax # Namespace where driver is installed
    # Used to specify configuration version
    configVersion: v3      # Refer the table containing the full list of supported drivers to find the appropriate config version
    replicas: 1
    forceUpdate: false     # Set to true in case you want to force an update of driver status
    common:                # All common specification
      image: "dellemc/csi-powermax:v1.4.0.000R"   #driver image for a particular release
      imagePullPolicy: IfNotPresent
          value: ""
        - name: X_CSI_K8S_CLUSTER_PREFIX
          value: "XYZ"

You can set the field replicas to a higher number than 1 for the latest driver versions.

Note - The image field should point to the correct image tag for version of the driver you are installing.
For e.g. - If you wish to install v2.7.0 of the CSI PowerMax driver, use the image tag dellemc/csi-powermax:v2.7.0


Although the sidecars field in the driver specification is optional, it is strongly recommended to not modify any details related to sidecars provided (if present) in the sample manifests. The only exception to this is modifications requested by the documentation, for example, filling in blank IPs or other such system-specific data. Any modifications not specifically requested by the documentation should be only done after consulting with Dell support.

Modify the driver specification

  • Choose the correct configVersion. Refer the table containing the full list of supported drivers and versions.
  • Provide the namespace (in metadata section) where you want to install the driver.
  • Provide a name (in metadata section) for the driver. This will be the name of the Custom Resource.
  • Edit the values for mandatory configuration parameters specific to your installation.
  • Edit/Add any values for optional configuration parameters to customize your installation.
  • If you are installing the latest versions of the CSI drivers, the default number of replicas is set to 2. You can increase/decrease this value.

StorageClass and VolumeSnapshotClass

New Installations

You should not provide any StorageClass or VolumeSnapshotClass details during driver installation. The sample files for all the drivers have been updated to reflect this change. Even if these details are there in the sample files, StorageClass or VolumeSnapshotClass will not be created.

What happens to my existing StorageClass & VolumeSnapshotClass objects

  • In case you are upgrading an existing driver installation by using kubectl edit or by patching the object in place, any existing objects will remain as is. If you added more objects as part of the upgrade, then this request will be ignored by the Operator.
  • If you uninstall the older driver, then any StorageClass or VolumeSnapshotClass objects will be deleted.
  • An uninstall and followed by an install of the driver would also result in StorageClass and VolumeSnapshotClass getting deleted and not getting created again.

NOTE: For more information on pre-requisites and parameters, please refer to the sub-pages below for each driver.

NOTE: Storage Classes and Volume Snapshot Classes would no longer be created during the installation of the driver via an operator from v1.4.0 and higher.

CSI to CSM Operator Migration

Migrating from CSI Operator to CSM Operator


Installing CSI Driver for PowerFlex via Operator


Installing CSI Driver for PowerMax via Operator


Installing CSI Driver for PowerScale via Operator


Installing CSI Driver for PowerStore via Operator

Unity XT

Installing CSI Driver for Unity XT via Operator