Help Center/ Cloud Container Engine/ User Guide/ Scheduling/ GPU Scheduling/ GPU Driver Version/ Configuring the GPU Driver Version Using a Node Label
Updated on 2026-06-16 GMT+08:00
View PDF
Share

Configuring the GPU Driver Version Using a Node Label

If you need to configure a differentiated driver for a node, you can add a specific label to the node and define a driver policy in the ConfigMap to configure a custom refined node-level driver.

Constraints

Ensure that CCE AI Suite (NVIDIA GPU) has been installed or upgraded to 2.13.1 or later.

Procedure

  1. Install or edit CCE AI Suite (NVIDIA GPU) using YAML.

    1. Log in to the CCE console and click the cluster name to access the cluster console.
    2. In the navigation pane, choose Add-ons. In the right pane, find CCE AI Suite (NVIDIA GPU) and click Edit. Click Edit YAML.

      If you install this add-on for the first time, find CCE AI Suite (NVIDIA GPU) in the right pane and click Install. Click Installation Using YAML.

    3. Edit the YAML file, add the following information to specify the node label name, and click Submit. 
      node_level_driver_label_name: <your-label-name>

      For example, set the label name to gpu.type.

  2. Configure the node label.

    1. In the navigation pane, choose Nodes. On the displayed page, click the Nodes tab.
    2. Locate the target node and choose More > Manage Labels in the Operation column.
    3. In the Manage Labels dialog box, click Add.
    4. Enter the label name (for example, gpu.type) set in 1 and a custom label value (for example, 535.216.03). Click OK.

  3. Create a custom driver download address for the specified label value.

    1. Use kubectl to access the cluster. For details, see Accessing a Cluster Using kubectl.
    2. Edit the ConfigMap in the kube-system namespace. 
      kubectl edit cm -n kube-system huawei-gpu-driver-config
    3. On the editing page, find the data field and add a new key-value pair in the following format: 
      custom_<label-value>: <driver-download-address>

      In this example, the label value is 535.216.03. For details about the driver download address, see CCE AI Suite (NVIDIA GPU).

    4. Save the modification and exit the editor.
    5. Check whether the ConfigMap has been updated. 
      kubectl get cm -n kube-system huawei-gpu-driver-config -o yaml

      Check whether the command output contains the custom_< label-value> field and the driver download address.

  4. Restart the node and check the new driver.

    1. Restart the node. 
      sudo reboot
    2. Log in to the node again after the node is restarted.
    3. Check whether the driver has been installed and is running normally. 
      nvidia-smi

      Confirm that the driver version displayed in the command output is the target version.

      +-----------------------------------------------------------------------------+
| NVIDIA-SMI 535.216.03   Driver Version: 535.216.03   CUDA Version: 12.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |

Parent Topic: GPU Driver Version