A container management platform has been deployed and is running properly, and its compatibility meets the requirements described in Kubernetes and OS Compatibility.
(Mandatory for enterprise storage) Initial configuration for interconnecting with Huawei enterprise storage has been completed, including storage pool division and port configuration. The version of the storage product meets the requirements in Compatibility with Huawei Enterprise Storage.
(Mandatory for distributed storage) Initial configuration for interconnecting with Huawei distributed storage has been completed, including storage pool division and port configuration. The version of the storage product meets the requirements in Compatibility with Huawei Distributed Storage.
The connectivity between Huawei storage and the container platform host has been configured. For example, the worker node running huawei-csi-controller communicates properly with the management IP address of the storage device to be connected, and the worker node running huawei-csi-node communicates properly with the service IP address of the storage device to be connected. In iSCSI scenarios, the ping command can be used to verify the connectivity.
Ensure that the language of the operating system is English.
Ensure that storage resource names, such as storage pool names and tenant names, are in English.

1.1 - Downloading the Huawei CSI Software Package

This section describes how to download the software package and the component structure of the software package.

Open a browser and enter https://github.com/Huawei/eSDK_K8S_Plugin/releases in the address box.
Download the software package of the 4.5.0 version based on the CPU architecture.
Software package naming rule: Plug-in name (eSDK_Huawei_Storage_Kubernetes_CSI_Plugin) + Version number + CPU architecture

Decompress the downloaded software package. The following table shows the component structure of the software package.

Table 1 Component description

Component	Description
image/huawei-csi-v 4.5.0-arch.tar	huawei-csi-driver image. arch is X86 or ARM.
image/storage-backend-controller-v 4.5.0-arch.tar	Back-end management controller image. arch is X86 or ARM.
image/storage-backend-sidecar-v 4.5.0-arch.tar	Back-end management sidecar image. arch is X86 or ARM.
image/huawei-csi-extender-v 4.5.0-arch.tar	huawei-csi-extender image. arch is X86 or ARM.
bin/	Binary file used by an image provided by Huawei.
bin/oceanctl	Command line tool provided by Huawei, which can be used to manage storage backends.
helm/	Helm project used to deploy Huawei CSI.
manual/	Used to manually install and deploy Huawei CSI.
examples/	.yaml sample file used during CSI use.
examples/backend	.yaml sample file used to create a storage backend.

1.2 - Uploading a Huawei CSI Image

Huawei provides the huawei-csi image for users. For details about how to obtain the image file, see Downloading the Huawei CSI Software Package.

To use the CSI image on the container management platform, you need to import the CSI image to the cluster in advance using either of the following methods:

(Recommended) Use Docker to upload the CSI image to the image repository.
Manually import the CSI image to all nodes where Huawei CSI needs to be deployed.

Uploading an Image to the Image Repository

The installation of Huawei CSI depends on the following image files provided by Huawei. Import and upload the image files in sequence. For details about how to obtain the image files, see Downloading the Huawei CSI Software Package.

huawei-csi-v4.5.0-arch.tar
storage-backend-controller-v4.5.0-arch.tar
storage-backend-sidecar-v4.5.0-arch.tar
huawei-csi-extender-v4.5.0-arch.tar

Prerequisites

A Linux host with Docker installed is available, and the host can access the image repository.

Procedure

Run the docker load -i huawei-csi-v4.5.0-arch.tar command to import the CSI image to the current node.
```
docker load -i huawei-csi-v4.5.0-arch.tar 
```
Run the docker tag huawei-csi:4.5.0 repo.huawei.com/huawei-csi:4.5.0 command to add the image repository address to the image tag. repo.huawei.com indicates the image repository address.
```
docker tag huawei-csi:4.5.0 repo.huawei.com/huawei-csi:4.5.0
```
Run the docker push repo.huawei.com/huawei-csi:4.5.0 command to upload the CSI image to the image repository. repo.huawei.com indicates the image repository address.
```
docker push repo.huawei.com/huawei-csi:4.5.0
```

You can also use containerd to import and upload the images.
For details about how to import and upload images to the CCE or CCE Agile platform, see the user manual of the platform.

Uploading an Image to a Local Node

If the image has been uploaded to the image repository, skip this section.

Prerequisites

The node has the corresponding Huawei CSI image file. For details about how to obtain the image file, see Downloading the Huawei CSI Software Package.
Docker or another container engine has been installed on the node.

Procedure

Use a remote access tool, such as PuTTY, to log in to the node where the image is to be imported through the management IP address.
Copy the image directory in the Kubernetes CSI component package to any directory on the current node.
Run the cd image command to go to the image working directory. For details about the tool path, see Table 1.
Run the following commands in sequence to import all Huawei CSI images in the image directory to the local node. In the commands, name indicates the name of a .tar image package.
Run the following command using the Docker container engine:
```
docker load -i <name>.tar
```
Run the following command using the containerd container engine:
```
ctr -n k8s.io image import <name>.tar
```
Run the following command using the Podman container engine:
```
podman load -i <name>.tar
```
If another container engine is installed on the node, use the image import command for the corresponding container engine.

1.3 - Checking User Configurations on Huawei Storage

After Huawei storage is connected to the container platform, Huawei CSI needs to manage storage resources on Huawei storage based on service requirements, such as creating and mapping volumes. In this case, Huawei CSI needs to use the users created on Huawei storage to communicate with Huawei storage. The following table lists the user information required for different storage devices.

Table 1 User requirements for connecting storage to CSI

Storage Type	User Type	Role	Level	Type
OceanStor V5	System user	Administrator	Administrator	Local user
OceanStor V5	vStore user	vStore administrator	Administrator	Local user
OceanStor Dorado V3	System user	Administrator	Administrator	Local user
OceanStor 6.1.3, 6.1.5, 6.1.6, 6.1.7, 6.1.8	System user	Administrator/User-defined role¹	N/A	Local user
OceanStor Dorado 6.1.0, 6.1.2, 6.1.3, 6.1.5, 6.1.6, 6.1.7, 6.1.8	System user	Administrator/User-defined role¹	N/A	Local user
	vStore user	vStore administrator	N/A	Local user
OceanStor Pacific series	System user	Administrator	N/A	Local user

Note 1: If a user-defined role is used, you need to configure permissions for the role. For details about how to configure the minimum permissions, see Configuring Custom Permissions.

You are advised not to use the users of the super administrator role.

1.4 - Checking Volume Snapshot-Dependent Components

This section describes how to check the volume snapshot-dependent components in the cluster.

Kubernetes earlier than v1.17.0 does not support the snapshot function. If the snapshot CRD is deployed, the cluster may be faulty. Therefore, if Huawei CSI is deployed on Kubernetes earlier than v1.17.0, perform the check according to Kubernetes Earlier Than v1.17.0.

Kubernetes Earlier Than v1.17.0

If the Kubernetes version is earlier than v1.17.0, the cluster may be faulty during snapshot deployment. Perform the following steps to delete the snapshot CRD installation file.

Run the following command to check the Kubernetes version. In the following example, the Kubernetes version is v1.16.0.

kubectl get node

The following is an example of the command output.

NAME          STATUS       ROLES     AGE    VERSION
test-master   Ready        master    311d   v1.16.0
test-node     Ready        <none>    311d   v1.16.0

Go to the /helm/esdk/crds/snapshot-crds directory and run the following command to delete the snapshot CRD installation file. For details about the component package path, see Table 1.
```
rm -rf ./huawei-csi-snapshot-crd-v1.yaml
```

1.5 - Checking the Host Multipathing Configuration

If you plan to use the FC/iSCSI/NVMe over RoCE/NVMe over FC protocol to access Huawei storage in a container environment, you are advised to use host multipathing software to enhance the link redundancy and performance of the host and storage. If you do not want to use the software, skip this section.

For details about the OSs and multipathing software supported by Huawei CSI, see Table 2.

If you want to use the FC/iSCSI protocol to connect to Huawei storage, you are advised to use native DM-Multipath provided by the OS.
If you want to use the NVMe over RoCE/NVMe over FC protocol to connect to Huawei storage, you are advised to use Huawei-developed UltraPath-NVMe.
If you want to use the SCSI protocol to connect to Huawei storage, disable DM-Multipath provided by the OS.

Prerequisites

Multipathing software has been correctly installed on a host.

If you use native DM-Multipath provided by the OS, contact your host or OS provider to obtain the documents and software packages required for the installation.
If you use Huawei-developed UltraPath or UltraPath-NVMe, contact Huawei engineers to obtain the UltraPath or UltraPath-NVMe documents and software packages. For details about the software package versions, see Table 1.

Procedure

If you use the iSCSI/FC protocol to connect to Huawei enterprise storage, configure and check host multipathing by referring to OceanStor Dorado and OceanStor Host Connectivity Guide for Red Hat.
If you use the NVMe over RoCE/NVMe over FC protocol to connect to Huawei enterprise storage, configure and check host multipathing by referring to OceanStor Dorado and OceanStor Host Connectivity Guide for Red Hat.
If you use iSCSI to connect to Huawei distributed storage, configure and check host multipathing by referring to Configuring Multipathing for an Application Server in FusionStorage 8.0.1 Block Storage Basic Service Configuration Guide.
If you use the native multipathing software provided by the OS, check whether the /etc/multipath.conf file contains the following configuration item.
```
defaults {
        user_friendly_names yes
        find_multipaths no
}
```
If the configuration item does not exist, add it to the beginning of the /etc/multipath.conf file.
For details about the functions of the user_friendly_names and find_multipaths parameters, see dm_multipath/config_file_defaults.

1.6 - Checking the Status of Host-Dependent Software

This section describes how to check whether the status of host-dependent software on worker nodes in a cluster is normal. In this example, the host OS is CentOS 7.9 x86_64.

Check the status of the iSCSI client.
```
systemctl status iscsi iscsid 
```
Check the status of the NFS client.
```
systemctl status rpcbind 
```

Check the status of DM-Multipath.

systemctl status multipathd.socket multipathd

Check the status of UltraPath.
```
systemctl status nxup
```
Check the status of UltraPath-NVMe.
```
systemctl status upudev upService_plus
```

1.7 - Checking the Images on Which CSI Depends

The installation of Huawei CSI depends on the images listed in the following table. If all worker nodes in the cluster have been connected to the Internet and can pull images online, you can skip this section. If nodes in the cluster cannot connect to the Internet, download the corresponding image file based on the Kubernetes version and upload it to the image repository or import it to all worker nodes in the Kubernetes cluster.

The huawei-csi-controller service depends on the following sidecar images: livenessprobe, csi-provisioner, csi-attacher, csi-resizer, csi-snapshotter, snapshot-controller, storage-backend-controller, storage-backend-sidecar, huawei-csi-driver, and huawei-csi-extender. The huawei-csi-node service depends on the following sidecar images: livenessprobe, csi-node-driver-registrar, and huawei-csi-driver.

For details about the functions and details of each image, see the following table.

Table 1 Images on which Huawei CSI depends

Container Name	Container Image	K8s Version Requirements	Feature Description
livenessprobe	k8s.gcr.io/sig-storage/livenessprobe:v2.5.0	v1.16+	This image is provided by the Kubernetes community, used to monitor the health status of CSI and report it to Kubernetes so that Kubernetes can automatically detect CSI program problems and restart the Pod to rectify the problems.
csi-resizer	k8s.gcr.io/sig-storage/csi-resizer:v1.4.0	v1.16+	This image is provided by the Kubernetes community, used to call CSI to provide more storage space for a PVC when expanding the capacity of the PVC.
csi-node-driver-registrar	k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0	v1.16+	This image is provided by the Kubernetes community, used to obtain CSI information and register a node with kubelet using the plug-in registration mechanism of kubelet so that Kubernetes can detect the connection between the node and Huawei storage.
csi-snapshotter	k8s.gcr.io/sig-storage/csi-snapshotter:v4.2.1	v1.17+	This image is provided by the Kubernetes community, used to call CSI to create or delete a snapshot on the storage system when creating or deleting a VolumeSnapshot.
snapshot-controller	k8s.gcr.io/sig-storage/snapshot-controller:v4.2.1	v1.17+	This image is provided by the Kubernetes community, used to listen to the VolumeSnapshot and VolumeSnapshotContent objects in the Kubernetes API and trigger csi-snapshotter to create a snapshot on the storage system when creating or deleting a VolumeSnapshot.
csi-provisioner	k8s.gcr.io/sig-storage/csi-provisioner:v3.0.0	v1.17+	This image is provided by the Kubernetes community, used to create or delete PVCs. Calls the huawei-csi-controller service to create a LUN or file system on the storage system as a PV when creating a PVC. Calls the huawei-csi-controller service to delete the LUN or file system corresponding to the PV when deleting a PVC.
csi-provisioner	quay.io/k8scsi/csi-provisioner:v1.4.0	v1.16.x
csi-attacher	k8s.gcr.io/sig-storage/csi-attacher:v3.4.0	v1.17+	Calls the huawei-csi-controller service to perform the "Publish/Unpublish Volume" operation when creating or deleting a Pod.
csi-attacher	quay.io/k8scsi/csi-attacher:v1.2.1	v.1.16.x
storage-backend-controller	storage-backend-controller:4.5.0	v1.16+	This image is provided by Huawei CSI software package, used to manage storageBackendClaim resources.
storage-backend-sidecar	storage-backend-sidecar:4.5.0	v1.16+	This image is provided by Huawei CSI software package, used to manage storageBackendContent resources.
huawei-csi-driver	huawei-csi:4.5.0	v1.16+	This image is provided by Huawei CSI software package, used to provide all features supported by Huawei CSI.
huawei-csi-extender	huawei-csi-extender:4.5.0	v1.16+	This image is provided by Huawei CSI software package, used to provide extended features of Huawei CSI.

If the cluster is not connected to the Internet, manually download the container images and upload them to the cluster. For details, see Downloading a Container Image.

2 - Installing Huawei CSI

This section describes how to install Huawei CSI.

In the current version, resource requests and limits are added to Huawei CSI. For details, see Huawei CSI Resource Management.

Prerequisites

Operations described in Installation Preparations have been completed.
All worker nodes of the cluster communicate properly with the service network of the storage device to be connected. In iSCSI scenarios, the ping command can be used to verify the connectivity.
Software clients required by the corresponding protocol, such as iSCSI and NFS clients, have been installed on all worker nodes of the cluster.

2.1 - Installing Huawei CSI Using Helm

Helm Installation Description

This section describes how to install Huawei CSI using Helm 3.

Huawei CSI can be installed as the root user or a non-root user. When installing Huawei CSI as a non-root user, ensure that the current user can access the API Server of the Kubernetes cluster. For details about how to configure access to the Kubernetes cluster as a non-root user, see Configuring Access to the Kubernetes Cluster as a Non-root User.
Huawei CSI must be run as the root user.

Helm is a software package management tool in the Kubernetes ecosystem. Similar to Ubuntu APT, CentOS YUM, or Python pip, Helm manages Kubernetes application resources.

You can use Helm to package, distribute, install, upgrade, and roll back Kubernetes applications in a unified manner.

For details about how to obtain and install Helm, see https://helm.sh/docs/intro/install/.
For details about the mapping between Helm and Kubernetes versions, see https://helm.sh/docs/topics/version_skew/.

When installing huawei-csi-controller, Helm deploys the following components in the workloads of the Deployment type in the specified namespace:

huawei-csi-driver: Huawei CSI driver.
storage-backend-controller: Huawei backend management controller, used to manage storageBackendClaim resources.
storage-backend-sidecar: used to manage storageBackendContent resources.
Kubernetes External Provisioner: used to provide or delete volumes.
Kubernetes External Attacher: used to attach or detach volumes.
Kubernetes External Resizer: used to expand the capacity of volumes.
Kubernetes External liveness-probe: used to determine the health status of a Pod.
(Optional) huawei-csi-extender: Huawei CSI extender.
(Optional) Kubernetes External Snapshotter: used to provide snapshot support (installed as CRD).
(Optional) Kubernetes External Snapshot Controller: used to control volume snapshots.

When installing huawei-csi-node, Helm deploys the following components in the workloads of the DaemonSet type in the specified namespace:

huawei-csi-driver: Huawei CSI driver.
Kubernetes Node Registrar: used to process driver registration.
liveness-probe: used to determine the health status of a Pod.

2.1.1 - Installing Huawei CSI on Kubernetes, OpenShift, and Tanzu

Installation Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the cluster through the management IP address.
Copy the helm directory in the Kubernetes CSI component package to any directory on the master node. For details about the Helm tool path, see Table 1.
Go to the helm/esdk working directory.
```
cd helm/esdk
```
Prepare the values.yaml file. Huawei CSI provides the values.yaml template file in the helm/esdk directory of the software package. You can also modify parameters according to Parameters in the values.yaml File of Helm to customize Huawei CSI.
Perform the following configuration before the installation:
- If the container platform is Kubernetes, skip this step.
- If the container platform is OpenShift, perform the configuration in Installation and Configuration on the OpenShift Platform.
- If the container platform is Tanzu, perform the configuration in Installation and Configuration on the Tanzu Platform.
Run the following command to update the storage backend CRD.
```
kubectl apply -f ./crds/backend/
```
(Optional) Check snapshot-dependent components by following the instructions provided in Checking Volume Snapshot-Dependent Components. After confirming that the components are correct, run the following command to update the snapshot CRD. If controller.snapshot.enabled is set to false or the Kubernetes version is earlier than v1.17, you can skip this step. For details, see Table 2.
```
kubectl apply -f ./crds/snapshot-crds/ --validate=false
```
Run the following command to install Huawei CSI. In the preceding command, helm-huawei-csi indicates the custom Helm chart name, ./ indicates that the Helm project in the current directory is used, and huawei-csi indicates the custom Helm chart namespace.
```
helm install helm-huawei-csi ./ -n huawei-csi --create-namespace
```
The following is an example of the command output.
```
NAME: helm-huawei-csi
LAST DEPLOYED: Wed Jun  8 11:50:28 2022
NAMESPACE: huawei-csi
STATUS: deployed
REVISION: 1
TEST SUITE: None
```

After the huawei-csi service is deployed, run the following command to check whether the service is started.

kubectl get pod -n huawei-csi

The following is an example of the command output. If the Pod status is Running, the installation is successful.

NAME                                     READY   STATUS    RESTARTS   AGE
huawei-csi-controller-6dfcc4b79f-9vjtq   9/9     Running   0          24m
huawei-csi-controller-6dfcc4b79f-csphc   9/9     Running   0          24m
huawei-csi-node-g6f4k                    3/3     Running   0          20m
huawei-csi-node-tqs87                    3/3     Running   0          20m

Installation and Configuration on the OpenShift Platform

For the OpenShift platform, run the following commands to create the SecurityContextConstraints resource.

Run the following command to edit the helm_scc.yaml file.
```
vi helm_scc.yaml
```

Modify the helm_scc.yaml file. In the following command output, huawei-csi indicates the created namespace. Replace it based on the actual situation.

apiVersion: security.openshift.io/v1
kind: SecurityContextConstraints
metadata:
  name: helm-scc
allowHostDirVolumePlugin: true
allowHostIPC: true
allowHostNetwork: true
allowHostPID: true
allowHostPorts: true
allowPrivilegeEscalation: true
allowPrivilegedContainer: true

defaultAddCapabilities:
- SYS_ADMIN
runAsUser:
  type: RunAsAny
seLinuxContext:
  type: RunAsAny
fsGroup:
  type: RunAsAny
users:
- system:serviceaccount:huawei-csi:huawei-csi-controller
- system:serviceaccount:huawei-csi:huawei-csi-node

Run the following command to create a SecurityContextConstraints file.
```
oc create -f helm_scc.yaml
```

Installation and Configuration on the Tanzu Platform

On the Tanzu platform, run the following command to configure the kubelet installation directory.

Go to the helm/esdk directory in the installation package, run the following command to open the configuration file, modify the file, and save the file. For details about the installation package directory, see Table 1.
```
vi values.yaml
```

Modify the kubeletConfigDir parameter as follows:

# Specify kubelet config dir path.
# kubernetes and openshift is usually /var/lib/kubelet
# Tanzu is usually /var/vcap/data/kubelet
# CCE is usually /mnt/paas/kubernetes/kubelet
kubeletConfigDir: /var/vcap/data/kubelet

For TKGI 1.16 or earlier of the Tanzu platform, run the following commands to configure the RBAC permission.

Run the following command to create a file named rbac.yaml.
```
vi rbac.yaml
```

Copy the following content to the rbac.yaml file, save the file, and exit.

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: huawei-csi-psp-role
rules:
- apiGroups: ['policy']
  resources: ['podsecuritypolicies']
  verbs: ['use']
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: huawei-csi-psp-role-cfg
roleRef:
  kind: ClusterRole
  name: huawei-csi-psp-role
  apiGroup: rbac.authorization.k8s.io
subjects:
- kind: Group
  apiGroup: rbac.authorization.k8s.io
  name: system:serviceaccounts:huawei-csi
- kind: Group
  apiGroup: rbac.authorization.k8s.io
  name: system:serviceaccounts:default

Run the following command to create the RBAC permission.
```
kubectl create -f rbac.yaml
```

2.1.2 - Installing Huawei CSI on the CCE or CCE Agile Platform

This section describes how to install Huawei CSI on the CCE or CCE Agile platform.

Creating a Helm Installation Package

The CCE or CCE Agile platform cannot directly install Huawei CSI using Helm. You need to manually create a Helm installation package and upload it to the chart list on the platform for installation.

Use a remote access tool, such as PuTTY, to log in to any node where Helm is deployed through the management IP address.
Copy the helm directory in the Huawei CSI component package to any directory on the node. For details about the Helm tool path, see Table 1.
Go to the helm working directory.
```
cd helm/
```

Modify the kubeletConfigDir and csiDriver.driverName parameters in the helm/esdk/values.yaml file.

vi ./esdk/values.yaml

Modify the following parameters:

# Specify kubelet config dir path.
# kubernetes and openshift is usually /var/lib/kubelet
# Tanzu is usually /var/vcap/data/kubelet
# CCE is usually /mnt/paas/kubernetes/kubelet
kubeletConfigDir: /mnt/paas/kubernetes/kubelet

# The CSI driver parameter configuration
csiDriver:
  # Driver name, it is strongly recommended not to modify this parameter
  # The CCE platform needs to modify this parameter, e.g. csi.oceanstor.com
  driverName: csi.oceanstor.com

Run the following command to create a Helm installation package. This command will generate the installation package to the current path.
```
helm package ./esdk/ -d ./
```

Installing Huawei CSI

Use a remote access tool, such as PuTTY, to log in to any master node where the CCE Agile platform is deployed through the management IP address.
Run the following command to create a namespace for deploying Huawei CSI. huawei-csi indicates the custom namespace.
```
kubectl create namespace huawei-csi
```
Export the Helm installation package. For details, see Creating a Helm Installation Package.
On the home page, choose Charts > My Charts > Upload Chart. The Upload Chart dialog box is displayed. Import the exported Helm installation package to the CCE Agile platform.
After the installation package is uploaded, choose Charts > My Charts. On the My Charts page that is displayed, choose Install > Submit. The chart release name can be customized.
On the home page, choose Charts > Releases and select the project specified during installation (for example, default in the following figure). After the installation is successful, Installed is displayed in the Status column.

2.1.3 - Parameters in the values.yaml File of Helm

When using Helm to install CSI, you need to prepare the values.yaml file of the Helm project based on the features required during deployment. Huawei CSI provides the values.yaml template file in the helm/esdk directory of the software package.

This section describes the configuration items in the values.yaml file and backend configuration examples in typical scenarios.

images Parameters

The images parameters in the values.yaml file are used to configure the component image information on which Huawei CSI depends during running. Set the following parameters:

Table 1 images parameters

Parameter	Description	Mandatory	Default Value
images.huaweiCSIService	huawei-csi image.	Yes	huawei-csi:4.5.0
images.storageBackendSidecar	Huawei back-end management sidecar image.	Yes	storage-backend-sidecar:4.5.0
images.storageBackendController	Huawei back-end management controller image.	Yes	storage-backend-controller:4.5.0
images.huaweiCSIExtender	huawei-csi-extender image.	No	huawei-csi-extender:4.5.0
images.sidecar.livenessProbe	livenessprobe sidecar image.	Yes	k8s.gcr.io/sig-storage/livenessprobe:v2.5.0
images.sidecar.provisioner	csi-provisioner sidecar image.	Yes	k8s.gcr.io/sig-storage/csi-provisioner:v3.0.0
images.sidecar.attacher	csi-attacher sidecar image.	Yes	k8s.gcr.io/sig-storage/csi-attacher:v3.4.0
images.sidecar.resizer	csi-resizer sidecar image.	Yes	k8s.gcr.io/sig-storage/csi-resizer:v1.4.0
images.sidecar.snapshotter	csi-snapshotter sidecar image.	Yes	k8s.gcr.io/sig-storage/csi-snapshotter:v4.2.1
images.sidecar.snapshotController	snapshot-controller sidecar image.	Yes	k8s.gcr.io/sig-storage/snapshot-controller:v4.2.1
images.sidecar.registrar	csi-node-driver-registrar sidecar image.	Yes	k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0

For details about the values of huaweiCSIService, storageBackendSidecar, storageBackendController, and huaweiCSIExtender, see Uploading a Huawei CSI Image. Use the name and version of the finally generated image.
For details about other sidecar image parameters, see Checking the Images on Which CSI Depends. Use the name and version of the finally uploaded image.

controller Parameters

The controller parameters are used to configure the huawei-csi-controller component.

Table 2 controller parameters

Parameter	Description	Mandatory	Default Value	Remarks
controller.controllerCount	Number of huawei-csi-controller component copies.	Yes	1	If the Kubernetes version is earlier than v1.17, the huawei-csi-controller component can be deployed only in single-copy mode because the csi-provisioner sidecar image provided by the Kubernetes community does not support the --leader-election parameter. Therefore, if the Kubernetes version is earlier than v1.17, this parameter can only be set to 1.
controller.volumeNamePrefix	PV name prefix. The default value is pvc, that is, the name of a created PV is pvc-<uuid>. The prefix must comply with the naming rules of a DNS subdomain name, and the total length of the PV name cannot exceed 253 characters.	No	pvc	The corresponding provisioner parameter name is --volume-name-prefix. It is recommended that the prefix contain no more than 20 characters. For details, see Configuring the PV Name Prefix. If the connected backend is OceanStor V5 SAN storage, it is recommended that the prefix contain a maximum of 5 characters. If the connected backend is OceanStor V5 NAS storage, the prefix can contain only lowercase letters, hyphens (-), and digits. If the connected backend is OceanStor Dorado or OceanStor storage, the prefix can contain only lowercase letters, hyphens (-), and digits. If the connected backend is OceanStor Pacific series storage, the prefix can contain a maximum of 58 characters, including only letters, digits, underscores (_), hyphens (-), and periods (.). If the connected backend is FusionStorage Block, the prefix can contain a maximum of 58 characters, including only letters, digits, underscores (_), and hyphens (-).
controller.webhookPort	Port used by the webhook service.	Yes	4433	If a port conflict occurs, change the port number to an idle one.
controller.snapshot.enabled	Whether to enable the snapshot feature.	Yes	true	If you want to use snapshot-related functions, enable this feature. The Kubernetes version must be later than v1.17.
controller.resizer.enabled	Whether to enable the capacity expansion feature.	Yes	true	The Kubernetes version must be later than v1.16.
controller.nodeSelector	Node selector of huawei-csi-controller. After this parameter is set, huawei-csi-controller will be scheduled only to a node with the label.	No	-	For details about the node selector, see Assign Pods to Nodes.
controller.tolerations	Taint toleration of huawei-csi-controller. After this parameter is set, huawei-csi-controller can tolerate taints on a node.	No	-	For details about taints and tolerations, see Taints and Tolerations.
controller.livenessProbePort	Liveness probe port of huawei-csi-controller, used for health check.	Yes	9808	If a port conflict occurs, change the port number to an idle one.
controller.csiExtender.volumeModify.enabled	Whether to enable the PVC change feature.	No	false	If you want to use PVC change-related functions, enable this feature.
controller.csiExtender.volumeModify.retryBaseDelay	Minimum retry interval when a PVC change fails to be created.	No	5s	The default value is recommended.
controller.csiExtender.volumeModify.retryMaxDelay	Maximum retry interval when a PVC change fails to be created.	No	5m	The default value is recommended.
controller.csiExtender.volumeModify.reconcileDelay	Interval for reconciling VolumeModifyClaim objects.	No	1s	The default value is recommended.

If controller.snapshot.enabled is set to true, you need to install the volume snapshot CRD resource in the helm/crd/snapshot-crds directory.

node Parameters

The node parameters are used to configure the huawei-csi-node component.

Table 3 node parameters

Parameter	Description	Mandatory	Default Value	Remarks
node.maxVolumesPerNode	Maximum number of volumes provisioned by Huawei CSI that can be used by a node. If this parameter is not specified or is set to 0, the number is unlimited. If nodeName is specified during Pod creation, this configuration will be ignored.	No	100	For details, see Volume Limits.
node.nodeSelector	Node selector of huawei-csi-node. After this parameter is set, huawei-csi-node will be scheduled only to a node with the label.	No	-	For details about the node selector, see Assign Pods to Nodes.
node.tolerations	Taint toleration of huawei-csi-node. After this parameter is set, huawei-csi-node can tolerate taints on a node.	No	- key: "node.kubernetes.io/memory-pressure" operator: "Exists" effect: "NoExecute" - key: "node.kubernetes.io/disk-pressure" operator: "Exists" effect: "NoExecute" - key: "node.kubernetes.io/network-unavailable" operator: "Exists" effect: "NoExecute"	For details about taints and tolerations, see Taints and Tolerations.
node.livenessProbePort	Liveness probe port of huawei-csi-node, used for health check.	Yes	9800	If a port conflict occurs, change the port number to an idle one.
node.kubeletVolumeDevicesDirName	Name of the directory where a block device is mounted to kubelet.	No	volumeDevices	After a block device is successfully mounted, the directory structure of the mount path is as follows: /var/lib/kubelet/plugins/kubernetes.io/csi/{kubeletVolumeDevicesDirName}/publish/{specName}/{podUID}

csiDriver Parameters

The csiDriver parameters include the basic configurations for running Huawei CSI, such as Huawei driver name and multipathing type.

Table 4 csiDriver parameters

Parameter	Description	Mandatory	Default Value	Remarks
csiDriver.driverName	Registered driver name.	Yes	csi.huawei.com	Use the default value. For the CCE Agile platform, modify this field. For example, csi.oceanstor.com.
csiDriver.endpoint	Communication endpoint.	Yes	/csi/csi.sock	Use the default value.
csiDriver.connectorThreads	Maximum number of disks that can be concurrently scanned/detached. The value is an integer ranging from 1 to 10.	Yes	4	A larger value indicates that more concurrent disk scanning and detaching operations are performed on a single node at the same time. When DM-Multipath is used, a large number of concurrent requests may cause unknown problems and affect the overall time.
csiDriver.volumeUseMultipath	Whether to use multipathing software. The value is a Boolean value.	Yes	true	It is strongly recommended that multipathing software be enabled to enhance the redundancy and performance of storage links.
csiDriver.scsiMultipathType	Multipathing software used when the storage protocol is fc or iscsi. The following parameter values can be configured: DM-multipath HW-UltraPath HW-UltraPath-NVMe	Mandatory when volumeUseMultipath is set to true.	DM-multipath	The DM-multipath value is recommended.
csiDriver.nvmeMultipathType	Multipathing software used when the storage protocol is roce or fc-nvme. Only HW-UltraPath-NVMe is supported.	Mandatory when volumeUseMultipath is set to true.	HW-UltraPath-NVMe	-
csiDriver.scanVolumeTimeout	Timeout interval for waiting for multipathing aggregation when DM-Multipath is used on the host. The value ranges from 1 to 600 seconds.	Yes	3	-
csiDriver.execCommandTimeout	Timeout interval for running commands on the host.	Yes	30	In scenarios such as mounting and capacity expansion, the CSI plug-in needs to run some host commands, for example, running the mount command to mount a file system. This parameter is used to control the timeout interval for running a single command.
csiDriver.allPathOnline	Whether to check whether the number of paths aggregated by DM-Multipath is equal to the actual number of online paths. The following parameter values can be configured: true: The drive letter mounting condition is met only when the number of paths aggregated by DM-Multipath is equal to the actual number of online paths. false: By default, the number of paths aggregated by DM-Multipath is not checked. As long as virtual drive letters are generated upon aggregation, the drive letter mounting condition is met.	This parameter is mandatory when csiDriver.scsiMultipathType is set to DM-multipath.	false	-
csiDriver.backendUpdateInterval	Interval for updating backend capabilities. The value ranges from 60 to 600 seconds.	Yes	60	-
csiDriver.controllerLogging.module	Record type of the controller log. The following parameter values can be configured: file console	Yes	file	When the value is file, logs are retained in the specified directory of the node. When the Pod where CSI is located is destroyed, logs are still retained. When the value is console, logs are retained in the temporary space of the Pod where CSI is located. When the Pod where CSI is located is destroyed, the logs are also destroyed.
csiDriver.controllerLogging.level	Output level of the controller log. The following parameter values can be configured: debug info warning error fatal	Yes	info	-
csiDriver.controllerLogging.fileDir	Directory of the controller log in file output mode.	Yes	/var/log/huawei	Ensure that the directory has sufficient space for storing logs. It is recommended that the space be greater than or equal to 200 MB.
csiDriver.controllerLogging.fileSize	Size of a single controller log file in file output mode.	Yes	20M	-
csiDriver.controllerLogging.maxBackups	Maximum number of controller log file backups in file output mode.	Yes	9	-
csiDriver.nodeLogging.module	Record type of the node log. The following parameter values can be configured: file console	Yes	file	When the value is file, logs are retained in the specified directory of the node. When the Pod where CSI is located is destroyed, logs are still retained. When the value is console, logs are retained in the temporary space of the Pod where CSI is located. When the Pod where CSI is located is destroyed, the logs are also destroyed.
csiDriver.nodeLogging.level	Output level of the node log. The following parameter values can be configured: debug info warning error fatal	Yes	info	-
csiDriver.nodeLogging.fileDir	Directory of the node log in file output mode.	Yes	/var/log/huawei	Ensure that the directory has sufficient space for storing logs. It is recommended that the space be greater than or equal to 200 MB.
csiDriver.nodeLogging.fileSize	Size of a single node log file in file output mode.	Yes	20M	-
csiDriver.nodeLogging.maxBackups	Maximum number of node log file backups in file output mode.	Yes	9	-

If Huawei CSI has been deployed in your container environment, ensure that the value of csiDriver.driverName is the same as that configured during previous deployment. Otherwise, existing volumes or snapshots provisioned by Huawei CSI in the system cannot be managed by the newly deployed Huawei CSI.

Other Parameters

Other parameters include some features of the CSI plug-in or the policies for obtaining images.

Table 5 Other parameters

Parameter	Description	Mandatory	Default Value	Remarks
kubernetes.namespace	Kubernetes namespace where Huawei CSI is running, which can be customized. The name must consist of lowercase letters, digits, and hyphens (-), for example, my-name and 123-abc.	No	huawei-csi	-
kubeletConfigDir	Working directory of kubelet.	Yes	/var/lib/kubelet	Use the default value. For the Tanzu platform, change the value of this field to /var/vcap/data/kubelet. For the CCE Agile platform, change the value of this field to /mnt/paas/kubernetes/kubelet.
sidecarImagePullPolicy	Pull policy of the sidecar image.	Yes	IfNotPresent	-
huaweiImagePullPolicy	Pull policy of the huawei-csi image.	Yes	IfNotPresent	-
CSIDriverObject.isCreate	Whether to create the CSIDriver object.	Yes	false	The CSIDriver feature is a GA version in Kubernetes v1.18. Therefore, to use this feature, the Kubernetes version must be later than v1.18. If the Kubernetes version is earlier than v1.18, set this parameter to false.
CSIDriverObject.attachRequired	Whether the CSI plug-in skips the attach operation. The following parameter values can be configured: true: The attach operation is required. false: The attach operation is skipped.	Yes	true	The attachRequired parameter can be configured in Kubernetes v1.18. If CSIDriverObject.isCreate is set to true and attachRequired is set to false, the huawei-csi plug-in will not deploy the csi-attacher sidecar. If NAS storage is used, this parameter can be set to false. If SAN storage is used, set this parameter to true.
CSIDriverObject.fsGroupPolicy	Whether the ownership and permissions of a basic volume can be changed before the volume is mounted. The following parameter values can be configured: "ReadWriteOnceWithFSType": The volume ownership and permission can be changed only when fsType is specified and accessModes of the volume contains ReadWriteOnce. "File": Kubernetes can use fsGroup to change the permissions and ownership of a volume to match fsGroup requested by a user in the Pod security policy, regardless of fsGroup or accessModes. "None": A volume is mounted without any change. "null": The fsGroupPolicy parameter is not set.	No	null	The fsGroupPolicy parameter can be configured in Kubernetes v1.20, and takes effect only when CSIDriverObject.isCreate is set to true. This feature is a Beta version in Kubernetes v1.20 but a GA version in Kubernetes v1.23. Therefore, the Kubernetes version must be later than v1.20.
leaderElection.leaseDuration	Leader duration.	No	8s	This parameter takes effect only in the multi-controller scenario.
leaderElection.renewDeadline	Time for the leader to be re-elected.	No	6s	This parameter takes effect only in the multi-controller scenario.
leaderElection.retryPeriod	Leader election retry time.	No	2s	This parameter takes effect only in the multi-controller scenario.

Ensure that the namespace entered in kubernetes.namespace exists on Kubernetes. If the namespace does not exist, run the following command to create it. In this example, the namespace for running Huawei CSI is huawei-csi.
kubectl create namespace huawei-csi

2.2 - Manually Installing Huawei CSI

This section describes how to manually install Huawei CSI.

Currently, only the Kubernetes platform supports manual installation of Huawei CSI.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the cluster through the management IP address.
Copy the manual directory in the Kubernetes CSI component package to any directory on the master node.
Run the following command to create a namespace.
```
kubectl create ns huawei-csi
```
Go to the manual/esdk working directory. For details about the path, see Table 1.
```
cd manual/esdk
```
Run the following command to update the storage backend CRD.
```
kubectl apply -f ./crds/backend/
```
(Optional) Check snapshot-dependent components by following the instructions provided in Checking Volume Snapshot-Dependent Components. After confirming that the components are correct, run the following command to update the snapshot CRD. If the Kubernetes version is earlier than v1.17, skip this step.
```
kubectl apply -f ./crds/snapshot-crds/ --validate=false
```
(Optional) Run the following command to install CSIDriver. If the CSIDriver feature is not used, you can skip this step. For details, see the CSIDriver feature.
```
kubectl apply -f ./deploy/csidriver.yaml 
```
Run the following command to install the huawei-csi-controller service.
If the Kubernetes version is earlier than v1.17, modify the ./deploy/huawei-csi-controller.yaml file as follows:
- If the Kubernetes version is earlier than v1.17, the snapshot feature is not supported. In this case, delete the snapshot-related container configurations items csi-snapshotter and snapshot-controller.
- If the Kubernetes version is earlier than v1.17, the csi-provisioner sidecar image provided by the Kubernetes community does not support the –leader-election parameter. Therefore, the leader-election parameter of the csi-provisioner container is deleted and only single-copy deployment is supported.
- Modify the dependent image version based on the version requirements in Checking the Images on Which CSI Depends.
```
kubectl apply -f ./deploy/huawei-csi-controller.yaml
```
Run the following command to install the huawei-csi-node service.
```
kubectl apply -f ./deploy/huawei-csi-node.yaml 
```

Run the following command to check whether the services are started.

kubectl get pod -n huawei-csi

The following is an example of the command output. If the Pod status is Running, the installation is successful.

NAME                                     READY   STATUS    RESTARTS   AGE
huawei-csi-controller-68745d489c-v5xkj   9/9     Running   0          13m
huawei-csi-node-4hbqp                    3/3     Running   0          13m
huawei-csi-node-f7dkf                    3/3     Running   0          13m
huawei-csi-node-xrntc                    3/3     Running   0          13m

In the multi-copy controller deployment scenario, you can modify the spec.replica field of the Deployment resource in the ./deploy/huawei-csi-controller.yaml file to specify the number of copies. After the modification, run the following command for the modification to take effect.
kubectl apply -f ./deploy/huawei-csi-controller.yaml

3 - Uninstalling Huawei CSI

This chapter describes how to uninstall Huawei CSI. The uninstallation method varies according to the installation mode.

If you do not uninstall Huawei CSI for the purpose of an upgrade, ensure that all resources (such as PV, PVC, snapshot, and storage backend resources) provisioned by Huawei CSI have been cleared on your container platform before uninstalling Huawei CSI. Otherwise, once you uninstall Huawei CSI, these resources cannot be automatically scheduled, managed, or cleared.

3.1 - Uninstalling Huawei CSI Using Helm

3.1.1 - Uninstalling Huawei CSI on Kubernetes, OpenShift, and Tanzu

This section describes how to uninstall Huawei CSI on the Kubernetes, OpenShift, and Tanzu platforms.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to uninstall Huawei CSI. In the command, helm-huawei-csi indicates the custom Helm chart name and huawei-csi indicates the namespace where the Helm chart resides. This command will uninstall the huawei-csi-controller, huawei-csi-node, and RBAC resources of Huawei CSI.
```
helm uninstall helm-huawei-csi -n huawei-csi
```
After the uninstallation command is executed, you need to check whether the uninstallation is successful. In the preceding command, huawei-csi indicates the namespace where the chart is located.
```
helm list -n huawei-csi
```
The following is an example of the command output. If the command output is empty, the service is successfully uninstalled.
```
NAME    NAMESPACE       REVISION        UPDATED STATUS  CHART   APP VERSION
```
Uninstall the huawei-csi-host-info object. For details, see Uninstalling the huawei-csi-host-info Object.
Uninstall the webhook resource. For details, see Uninstalling a Webhook Resource.
(Optional) Uninstall the snapshot-dependent component service. For details, see Uninstalling the Snapshot-Dependent Component Service.
(Optional) Uninstall the Lease resource. For details, see Uninstalling a Lease Resource.

3.1.2 - Uninstalling Huawei CSI on CCE or CCE Agile

This section describes how to uninstall Huawei CSI on the CCE or CCE Agile platform. The following uses CCE Agile v22.3.2 as an example.

Procedure

Log in to the CCE Agile platform.
On the home page, choose Charts > Releases. The Releases page is displayed.
Select a Huawei CSI release and click Uninstall. In the displayed dialog box, click OK.
Uninstall the huawei-csi-host-info object. For details, see Uninstalling the huawei-csi-host-info Object.
Uninstall the webhook resource. For details, see Uninstalling a Webhook Resource.
(Optional) Uninstall the snapshot-dependent component service. For details, see Uninstalling the Snapshot-Dependent Component Service.

3.1.3 - Uninstalling CSI-Dependent Component Services

This section describes how to uninstall the CSI-dependent component services.

Uninstalling the huawei-csi-host-info Object

Secret object huawei-csi-host-info stores the initiator information about each node in the cluster, for example, iSCSI initiators. When you run the helm uninstall command, the resource will not be uninstalled. To uninstall the resource, perform the following steps:

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to delete the Secret object. huawei-csi-host-info is the name of the Secret object, and huawei-csi is the namespace where the Secret object is located.
```
kubectl delete secret huawei-csi-host-info -n huawei-csi
```
Run the following command to check whether the Secret object is successfully uninstalled.
```
kubectl get secret huawei-csi-host-info -n huawei-csi 
```
The following is an example of the command output. If NotFound is displayed in the command output, the huawei-csi-host-info object is successfully uninstalled.
```
Error from server (NotFound): secrets "huawei-csi-host-info" not found
```

Uninstalling a Webhook Resource

The webhook resource named storage-backend-controller.xuanwu.huawei.io is used to verify the backend key information and connectivity with the storage. When you run the helm uninstall command, the resource will not be uninstalled. To uninstall the resource, perform the following steps:

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.

Run the following command to query the webhook-dependent component service.

kubectl get validatingwebhookconfigurations.admissionregistration.k8s.io storage-backend-controller.xuanwu.huawei.io

The following is an example of the command output.

NAME                                          WEBHOOKS   AGE
storage-backend-controller.xuanwu.huawei.io   1          12d

Run the following command to uninstall the webhook-dependent component service.

kubectl delete validatingwebhookconfigurations.admissionregistration.k8s.io storage-backend-controller.xuanwu.huawei.io

Run the following command to check whether the service is successfully uninstalled. If the command output is empty, the uninstallation is successful.
```
kubectl get validatingwebhookconfigurations.admissionregistration.k8s.io storage-backend-controller.xuanwu.huawei.io
```

Uninstalling the Snapshot-Dependent Component Service

Do not uninstall the snapshot-dependent component service when snapshots exist. Otherwise, Kubernetes will automatically delete all user snapshots and they cannot be restored. Exercise caution when performing this operation. For details, see Delete a CustomResourceDefinition.
Do not uninstall the snapshot-dependent component service during the CSI upgrade.

Scenario Description

Currently, Huawei CSI uses the snapshot feature.
Currently, only Huawei CSI is available in the Kubernetes cluster, and Huawei CSI is no longer used.
Before the uninstallation, ensure that no VolumeSnapshot resource managed by Huawei CSI exists in the Kubernetes cluster.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.

Run the following command to uninstall the snapshot-dependent component service.

kubectl delete crd volumesnapshotclasses.snapshot.storage.k8s.io volumesnapshotcontents.snapshot.storage.k8s.io volumesnapshots.snapshot.storage.k8s.io

Run the following command to check whether the service is successfully uninstalled. If the command output is empty, the uninstallation is successful.
```
kubectl get crd | grep snapshot.storage.k8s.io
```

Uninstalling a Lease Resource

If the value of the controller.controllerCount configuration item in the values.yaml file is greater than 1, huawei-csi-controller will be deployed in multi-copy mode. The multiple copies of huawei-csi-controller are implemented using the LeaderElection mechanism of Kubernetes. This mechanism creates a Lease object to store the current Holder information. When you run the helm uninstall command, the resource will not be uninstalled. To uninstall the resource, perform the following steps. If the value of controller.controllerCount is 1, you can skip the following steps. For details about the configuration item, see Table 2.

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.

Run the following command to query the Lease information.

kubectl get lease -n huawei-csi

The following is an example of the command output.

NAME                                         HOLDER    AGE
csi-huawei-com                               node-1    24d
external-attacher-leader-csi-huawei-com      node-1    24d
external-resizer-csi-huawei-com              node-1    24d
external-snapshotter-leader-csi-huawei-com   node-1    24d
snapshot-controller-leader                   node-1    24d
storage-backend-controller                   node-1    24d
huawei-csi-extender                          node-1    24d

Run the following command to uninstall the Lease resource.

kubectl delete lease -n huawei-csi csi-huawei-com external-attacher-leader-csi-huawei-com external-resizer-csi-huawei-com external-snapshotter-leader-csi-

Run the following command to check whether the uninstallation is successful.
```
kubectl get lease -n huawei-csi
```
The following is an example of the command output. If the command output is empty, the uninstallation is successful.
```
No resources found in huawei-csi namespace.
```

3.2 - Manually Uninstalling Huawei CSI

This section describes how to manually uninstall Huawei CSI.

If you do not uninstall Huawei CSI for the purpose of an upgrade, ensure that all resources (such as PV, PVC, snapshot, and storage backend resources) provisioned by Huawei CSI have been cleared on your container platform before uninstalling Huawei CSI. Otherwise, once you uninstall Huawei CSI, these resources cannot be automatically scheduled, managed, or cleared.

Uninstalling the huawei-csi-node Service

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to uninstall the huawei-csi-node service. Replace huawei-csi with the namespace where Huawei CSI is located.
```
kubectl delete daemonset huawei-csi-node -n huawei-csi
```
Run the following command to check whether the service is successfully uninstalled. If NotFound is displayed, the service is successfully uninstalled.
```
kubectl get daemonset huawei-csi-node -n huawei-csi
```

Uninstalling the huawei-csi-controller Service

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to uninstall the huawei-csi-controller service. Replace huawei-csi with the namespace where Huawei CSI is located.
```
kubectl delete deployment huawei-csi-controller -n huawei-csi
```
Run the following command to check whether the service is successfully uninstalled. If NotFound is displayed, the service is successfully uninstalled.
```
kubectl get deployment huawei-csi-controller -n huawei-csi
```

Uninstalling the csidriver Object

If the CSIDriver feature is not used during installation, you can skip the following steps.

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to uninstall the csidriver object.
```
kubectl delete csidriver csi.huawei.com
```
Run the following command to check whether the service is successfully uninstalled. If NotFound is displayed, the service is successfully uninstalled.
```
kubectl get csidriver csi.huawei.com
```

Deleting the RBAC Permission

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.

Delete the RBAC permission.

kubectl -n huawei-csi -l provisioner=csi.huawei.com delete ServiceAccount,Service,role,rolebinding,ClusterRole,ClusterRoleBinding

Uninstalling Other Resources

Uninstall the huawei-csi-host-info object. For details, see Uninstalling the huawei-csi-host-info Object.
Uninstall the webhook resource. For details, see Uninstalling a Webhook Resource.
(Optional) Uninstall the snapshot-dependent component service. For details, see Uninstalling the Snapshot-Dependent Component Service.
(Optional) Uninstall the Lease resource. For details, see Uninstalling a Lease Resource.

4 - Upgrading or Rolling Back Huawei CSI

This section describes how to upgrade or roll back Huawei CSI.

In the current version, resource requests and limits are added to Huawei CSI. For details, see Huawei CSI Resource Management.

4.1 - Upgrading or Rolling Back Huawei CSI Using Helm

To upgrade Huawei CSI from 2.x to 4.5.0, uninstall it by referring to the user guide of the earlier version and install Huawei CSI by referring to Installing Huawei CSI Using Helm.

To upgrade Huawei CSI from 2.x or 3.x to 4.5.0, see Upgrading from 2.x or 3.x to 4.x.

To upgrade Huawei CSI from 4.x to 4.5.0, see Upgrading Huawei CSI on Kubernetes, OpenShift, and Tanzu.

4.1.1 - Rolling Back Huawei CSI on Kubernetes, OpenShift, and Tanzu

Prerequisites

CSI has been updated using Helm 3.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Go to the helm/esdk working directory. For the directory path, see Table 1.
```
cd helm/esdk
```

Run the following command to query the historical versions of the CSI services deployed using Helm.

helm history helm-huawei-csi -n huawei-csi

The following is an example of the command output.

REVISION        UPDATED                         STATUS          CHART           APP VERSION     DESCRIPTION     
1       	Mon Jan  8 04:15:40 2024	superseded	esdk-4.4.0	4.4.0      	Install complete
2       	Mon Jan  8 04:16:12 2024	deployed  	esdk-4.5.0	4.5.0      	Upgrade complete

Run the following command to roll back the CSI services to the specified version.
In the preceding command, revision-number indicates a version number queried in 3. For example, the version is 1.
```
helm rollback helm-huawei-csi -n huawei-csi 1
```
The following is an example of the command output. If Rollback was a success is displayed in the command output, the CSI services are successfully rolled back to the specified version.
```
Rollback was a success! Happy Helming!
```

4.1.2 - Upgrading Huawei CSI

This section describes how to upgrade Huawei CSI.

During the upgrade or rollback, the existing resources such as PVCs, snapshots, and Pods will run properly and will not affect your service access.

Some CSI 2.x versions are unavailable now. If the upgrade fails, CSI may fail to be rolled back to a version which is unavailable now.
After an upgrade from 2.x, 3.x, or 4.x to 4.5.0, a Pod that has been provisioned in the source version may fail to be mounted again. For details, see Upgrading from 2.x or 3.x to 4.x.
During the upgrade or rollback, you cannot use Huawei CSI to create new resources or mount or unmount an existing PVC.
During the upgrade or rollback, do not uninstall the snapshot-dependent component service.

4.1.2.1 - Upgrading from 2.x or 3.x to 4.x

In CSI 2.x or 3.x, when block storage is used, the mapping with storage is set up in the huawei-csi-node service. Therefore, the huawei-csi-node service needs to communicate with the storage management network. Because the huawei-csi-node service is deployed as a DaemonSet, the huawei-csi-node service is deployed on each node in the cluster. As a result, in a large-scale cluster, each huawei-csi-node service sends requests to the storage and the number of storage connections may be fully occupied. Accordingly, huawei-csi-node cannot provide services properly. In CSI 4.x, the deployment model is optimized. The setup of the mapping with storage is migrated to the huawei-csi-controller service and the huawei-csi-node service does not need to communicate with the storage management network. This reduces the networking complexity of Huawei CSI. In addition, the huawei-csi-controller service is deployed as a Deployment. The number of copies is set based on the customer’s reliability requirements. Generally, the number of copies ranges from 1 to 3. Therefore, the number of connections between Huawei CSI and storage is greatly reduced, so that Huawei CSI can connect to a large-scale cluster. This change may cause a problem. That is, if a new mount process is generated after CSI is upgraded to 4.x but with workloads provisioned using 2.x or 3.x and the Container Orchestration (CO) system does not invoke the huawei-csi-controller service provided by Huawei CSI, the mounting will fail. For details, see A Pod Fails to Be Created and Message “publishInfo doesn’t exist” Is Displayed in the Events Log.

Backing Up Storage Backend Configurations

If you have evaluated the risks mentioned in the preceding notice and need to upgrade CSI from 2.x or 3.x to 4.5.0, perform the following steps to back up storage backend configurations:

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to back up the backend information to the configmap.json file. For the OpenShift platform, replace kubectl with oc.
```
kubectl get cm huawei-csi-configmap -n huawei-csi -o json > configmap.json
```

Upgrading Huawei CSI

Perform the upgrade according to the procedure described in Upgrading Huawei CSI.

Configuring the Storage Backend

Configure the storage backend by following the instructions in Managing Storage Backends according to the backend information backed up in Backing Up Storage Backend Configurations. After the storage backend is successfully configured, perform operations according to the risk handling methods described in the preceding notice to prevent problems during Pod failover.

4.1.2.2 - Upgrading Huawei CSI on Kubernetes, OpenShift, and Tanzu

Prerequisites

Huawei CSI of an earlier version is installed using Helm.
A Huawei CSI image of a new version has been created and uploaded to the image repository or imported to all nodes by following the instructions provided in Uploading a Huawei CSI Image.

Upgrading Huawei CSI

If CSI of an earlier version is deployed using Helm, perform the following steps to upgrade Huawei CSI.

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Copy the CSI component package of the target version to any directory on the master node.
Go to the helm/esdk working directory. For the directory path, see Table 1.
```
cd helm/esdk
```
Run the kubectl apply -f ./crds/backend/ command to update the storage backend CRD.
```
kubectl apply -f ./crds/backend/
```
(Optional) Check snapshot-dependent components by following the instructions provided in Checking Volume Snapshot-Dependent Components. After confirming that the components are correct, run the kubectl apply -f ./crds/snapshot-crds/ –validate=false command to update the snapshot CRD. If controller.snapshot.enabled is set to false or the Kubernetes version is earlier than v1.17, you can skip this step. For details, see Table 2.
```
kubectl apply -f ./crds/snapshot-crds/ --validate=false
```
Run the following command to obtain the original service configuration file. helm-huawei-csi indicates the Helm chart name specified during the installation of the earlier version, and huawei-csi indicates the Helm chart namespace specified during the installation of the earlier version.
```
helm get values helm-huawei-csi -n huawei-csi -a > ./update-values.yaml
```

Run the vi update-values.yaml command to open the file obtained in 6, modify the images configuration items, and update the image to the latest version. For details about the parameters to be modified, see Table 1.

Table 1 images configuration items

Parameter	Description	New Value
images.huaweiCSIService	huawei-csi image.	huawei-csi:4.5.0
images.storageBackendSidecar	Image used by Huawei backends to manage storageBackendContent resources.	storage-backend-sidecar:4.5.0
images.storageBackendController	Image used by Huawei backends to manage storageBackendClaim resources.	storage-backend-controller:4.5.0
images.huaweiCSIExtender	huawei-csi-extender image.	huawei-csi-extender:4.5.0
images.sidecar.livenessProbe	livenessprobe sidecar image.	k8s.gcr.io/sig-storage/livenessprobe:v2.5.0
images.sidecar.provisioner	csi-provisioner sidecar image.	k8s.gcr.io/sig-storage/csi-provisioner:v3.0.0
images.sidecar.attacher	csi-attacher sidecar image.	k8s.gcr.io/sig-storage/csi-attacher:v3.4.0
images.sidecar.resizer	csi-resizer sidecar image.	k8s.gcr.io/sig-storage/csi-resizer:v1.4.0
images.sidecar.snapshotter	csi-snapshotter sidecar image.	k8s.gcr.io/sig-storage/csi-snapshotter:v4.2.1
images.sidecar.snapshotController	snapshot-controller sidecar image.	k8s.gcr.io/sig-storage/snapshot-controller:v4.2.1
images.sidecar.registrar	csi-node-driver-registrar sidecar image.	k8s.gcr.io/sig-storage/csi-node-driver-registrar:v2.3.0

(Optional) If you need to update configuration items or add configuration information during the upgrade, modify the configuration information in the update-values.yaml file by referring to Parameters in the values.yaml File of Helm.
During the upgrade, if the update-values.yaml and values.yaml configuration files contain the same configuration item, the configuration in the update-values.yaml file takes effect preferentially.
Run the following command to upgrade Huawei CSI. In the following command, helm-huawei-csi indicates the specified Helm chart name, huawei-csi indicates the specified Helm chart namespace, and update-values.yaml indicates the file obtained in 6.
```
helm upgrade helm-huawei-csi ./ -n huawei-csi -f ./values.yaml -f ./update-values.yaml
```

After the huawei-csi service is deployed, run the following command to check whether the service is started.

kubectl get pod -n huawei-csi

The following is an example of the command output. If the Pod status is Running, the service is started successfully.

NAME                                     READY   STATUS    RESTARTS   AGE
huawei-csi-controller-6dfcc4b79f-9vjtq   9/9     Running   0          24m
huawei-csi-controller-6dfcc4b79f-csphc   9/9     Running   0          24m
huawei-csi-node-g6f4k                    3/3     Running   0          20m
huawei-csi-node-tqs87                    3/3     Running   0          20m

4.1.2.3 - Upgrading Huawei CSI on CCE or CCE Agile

Prerequisites

You have downloaded the CSI software package of a new version.

Procedure

Uninstall CSI. For details, see Uninstalling Huawei CSI on CCE or CCE Agile.
Install CSI of the new version. For details, see Installing Huawei CSI on the CCE or CCE Agile Platform.

4.1.3 - Rolling Back Huawei CSI on CCE or CCE Agile

During the upgrade or rollback, the existing resources such as PVCs, snapshots, and Pods will run properly and will not affect your service access.
During the upgrade or rollback, you cannot use Huawei CSI to create new resources or mount or unmount an existing PVC.
During the upgrade or rollback, do not uninstall the snapshot-dependent component service.

Prerequisites

You have downloaded the CSI software package of the source version.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Uninstall CSI. For details, see Procedure.
Reinstall CSI of the source version. For details, see Installing Huawei CSI on the CCE or CCE Agile Platform.

4.2 - Manual Upgrade/Rollback

4.2.1 - Upgrading Huawei CSI

This section describes how to manually upgrade Huawei CSI.

During the upgrade or rollback, the existing resources such as PVCs, snapshots, and Pods will run properly and will not affect your service access.

Some CSI 2.x versions are unavailable now. If the upgrade fails, CSI may fail to be rolled back to a version which is unavailable now.
During the upgrade or rollback, you cannot use Huawei CSI to create new resources or mount or unmount an existing PVC.
During the upgrade or rollback, do not uninstall the snapshot-dependent component service.

Upgrading CSI from 2.x or 3.x to 4.5.0

To upgrade CSI from 2.x or 3.x to 4.5.0, perform the following operations:

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Run the following command to back up the backend information to the configmap.json file. For the OpenShift platform, replace kubectl with oc.
```
kubectl get cm huawei-csi-configmap -n huawei-csi -o json > configmap.json
```
Uninstall CSI. For details, see Manually Uninstalling Huawei CSI.
Install CSI of the current version. For details, see Manually Installing Huawei CSI.
Install the backend information backed up in 2 according to Managing Storage Backends.

Upgrading CSI from 4.x to 4.5.0

To upgrade CSI from 4.x to 4.5.0, perform the following operations:

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Uninstall CSI. For details, see Manually Uninstalling Huawei CSI.
Install CSI of the current version. For details, see Manually Installing Huawei CSI.

4.2.2 - Rolling Back Huawei CSI

Uninstall CSI by referring to Manually Uninstalling Huawei CSI, and then download and install CSI of the source version.

During the upgrade or rollback, the existing resources such as PVCs, snapshots, and Pods will run properly and will not affect your service access.
During the upgrade or rollback, you cannot use Huawei CSI to create new resources or mount or unmount an existing PVC.
During the upgrade or rollback, do not uninstall the snapshot-dependent component service.

Prerequisites

You have downloaded the CSI software package of the source version.

Procedure

Use a remote access tool, such as PuTTY, to log in to any master node in the Kubernetes cluster through the management IP address.
Uninstall CSI. For details, see Manually Uninstalling Huawei CSI.
Reinstall CSI of the source version. For details, see Manually Installing Huawei CSI.