Container Service for Kubernetes NVIDIA GPUs accelerate compute-intensive workloads such as scientific computing and graphics rendering. Container Service for Kubernetes provides unified scheduling and O&M for compute-optimized GPU resources of various models, improving GPU cluster utilization. This topic explains how to add GPU nodes to a cluster.
Prerequisites
An ACK Pro cluster or ACK dedicated cluster (no longer available for creation) is created.
Create a GPU node pool
-
Log on to the ACK console. In the left navigation pane, click Clusters.
-
On the Clusters page, click the name of your cluster. In the left navigation pane, click .
-
Click Create Node Pool. Set Instance Type to Elastic GPU Service and Expected Nodes to your desired count. See Create and manage a node pool for other parameters.
See GPU instance types supported by ACK.
Note-
If no instance types are available, select a different vSwitch.
-
On Ubuntu 22.04 or RHEL 9.3 64-bit nodes, the NVIDIA Device Plugin sets
NVIDIA_VISIBLE_DEVICES=allfor pods by default. Runningsystemctl daemon-reloadorsystemctl daemon-reexecon the node may cause GPU device access loss, stopping the NVIDIA Device Plugin. See What do I do if the "Failed to initialize NVML: Unknown Error" error occurs when I run a GPU container?.
-
View attached GPU devices
After creating the node pool, view the GPU devices attached to each node.
-
Log on to the ACK console. In the left navigation pane, click Clusters.
-
On the Clusters page, click the name of your cluster. In the left navigation pane, click .
-
Click Details in the Actions column to view attached GPU devices. To find the GPU model, click the ECS instance ID to open the instance details page. On the Instance Details tab, check the GPU field under Configuration Information. For example,
GPU: NVIDIA T4forecs.gn6i-c4g1.xlarge.