Installation Preparations

• 1: Downloading the Huawei CSI Software Package
• 2: Uploading a Huawei CSI Image
• 3: Checking User Configurations on Huawei Storage
• 4: Checking Volume Snapshot-Dependent Components
• 5: Checking the Host Multipathing Configuration
• 6: Checking the Status of Host-Dependent Software
• 7: Checking the Images on Which CSI Depends

1 - Downloading the Huawei CSI Software Package

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

1. Open a browser and enter https://github.com/Huawei/eSDK_K8S_Plugin/releases in the address box.

2. 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

3. 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-v4.5.0-arch.tar	huawei-csi-driver image. arch is X86 or ARM.
   image/storage-backend-controller-v4.5.0-arch.tar	Back-end management controller image. arch is X86 or ARM.
   image/storage-backend-sidecar-v4.5.0-arch.tar	Back-end management sidecar image. arch is X86 or ARM.
   image/huawei-csi-extender-v4.5.0-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.

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

1. 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 
   

2. 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
   

3. 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

1. 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.

2. Copy the image directory in the Kubernetes CSI component package to any directory on the current node.

3. Run the cd image command to go to the image working directory. For details about the tool path, see Table 1.

4. 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.

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

• 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.

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.

1. 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
   

2. 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
   

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

1. 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.

2. 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.

3. 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.

4. 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.

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
  

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

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.