Using Ascend GPU in the Application¶

This section explains how to use Ascend GPU on the DCE 5.0 platform.

Prerequisites¶

• The Ascend GPU driver has been installed on the current cluster.
• The GPU cards in the current cluster have not undergone any virtualization operations or been occupied by other applications.

Configuration through the User Interface¶

1. Confirm whether the cluster has detected the GPU cards. Click Clusters -> Cluster Settings -> Addon Plugins and check if the corresponding GPU type has been automatically enabled and detected. Currently, the cluster will automatically enable GPU , and set the GPU type as Ascend .

2. Deploy the workload by clicking on Clusters -> Workloads . Deploy the workload using an image, and after selecting the type (Ascend), configure the number of physical cards that the application will use:

Number of Physical Cards (huawei.com/Ascend910): Indicates the number of physical cards that the current Pod needs to mount. The input value must be an integer and less than or equal to the number of cards on the host machine.

If there are issues with the above configuration, scheduling failures and resource allocation problems may occur.

YAML Configuration¶

To apply for GPU resources in the workload, add the huawei.com/Ascend910 parameter in the resource request and limit configuration to specify the physical card resources used by the application.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: full-Ascend-gpu-demo
  namespace: default
spec:
  replicas: 1
  selector:
    matchLabels:
      app: full-Ascend-gpu-demo
  template:
    metadata:
      labels:
        app: full-Ascend-gpu-demo
    spec:
      containers:
      - image: nginx:perl
        name: container-0
        resources:
            limits:
              cpu: 250m
              huawei.com/Ascend910: '1'
              memory: 512Mi
            requests:
              cpu: 250m
              memory: 512Mi
      imagePullSecrets:
      - name: default-secret

Comments