Docs

Volcengine Kubernetes Engine

NVIDIA GPU

Configuring NVIDIA GPU scheduling

NVIDIA GPU scheduling is the basic scheduling mode in VKE that allows a container to occupy one or more entire GPUs with full access to all computing power and video memory of these GPUs. This topic describes how to configure NVIDIA GPU scheduling for a cluster.

Prerequisites

The nvidia-device-plugin add-on has been installed. For details, see Installing add-ons.

Procedure

Step 1: Create a GPU node pool

In the left navigation bar, click Cluster. In the cluster list, click the name of the target cluster to open the cluster management page.

In the left navigation bar, click Node pool.

On the Node pool page, click Create node pool to create a node pool that meets your requirements. For parameter details, see Creating node pools.

For Instance type, select GPU Accelerated to add nodes with GPUs to the cluster. For specification details, see GPU Accelerated.

(Optional) In the Advanced settings section, click Add new to add a GPU scheduling label to the node pool. In this example, set the key to vke.node.gpu.schedule and the value to nvidia. This setting enables NVIDIA GPU scheduling for nodes in this node pool.

tip

For more information about the tag for GPU scheduling, see GPU scheduling overview.

After configuring other parameters, click Confirm to create the node pool.

Step 2: Configure a workload

This topic uses the creation of a stateless workload (Deployment) as an example. For detailed steps, see Creating Deployments.

Creating through the console

In the left navigation bar of the cluster management page, click Workload > Deployment.

Create a Deployment or update an existing Deployment. In the ResourceQuota section of the Container configuration step, configure GPU resources as needed.

Parameter	Description
GPU resource type	In this example, select Nvidia.
GPU computing power	Configure the number of whole GPUs requested by the Pod. The value must be a positive integer. For example, if the value is 1, it means requesting one whole GPU.
Graphics card type	Configure the GPU nodes to which the pod is to be scheduled. By default, the GPU model is not limited. After you specify the GPU model, GPU nodes will be specified based on the selected model.

tip

The parameters displayed in the console vary based on whether the mgpu add-on is installed in the cluster where the workload resides.

If the mgpu add-on is not installed, the GPU computing power and Graphics card type parameters are displayed. Specify the parameters based on your business requirements.

If the mgpu add-on is installed, the GPU computing power and GPU memory parameters are displayed. In the NVIDIA GPU scheduling scenario, you need to set the GPU computing power parameter to an integer greater than or equal to 1. In this case, GPU memory will change to Graphics card type, and the system will display Enable NVIDIA scheduling.

In the NVIDIA GPU scheduling scenario, you need to select Enable NVIDIA Scheduling. If you do not select this option, NVIDIA GPU scheduling will not be used, and mGPU scheduling will be used instead. Both mGPU scheduling and the mGPU add-on are public beta features.

Creating through kubectl

Use YAML to create a workload that applies NVIDIA GPU scheduling:

Validating results

Checking GPU status

Run the following command to log in to the container and view the total allocated video memory.

The expected result is as follows. You can see information such as the GPU type and driver used by the Pod.

Viewing GPU monitoring data

After NVIDIA GPU scheduling is enabled, you can view the GPU resource dashboard in the AI resource observability module of cloud-native observability. For details, see AI resource observability.

Last updated: 2026.07.24 14:51:38