9.18.11.3.2. Labeling nodes with a MIG-backed vGPU profile
If you have GPUs that support NVIDIA Multi-Instance GPU (MIG), you can select a MIG-backed vGPU instance instead of time-sliced vGPU instances. When you use MIG, you give a partition of dedicated hardware to selected VMs.
Prerequisites
- You have configured vGPU support. For more information see MIG Support in OpenShift Container Platform.
- You have the NVIDIA GPU Operator version 25.10 or higher.
- You are using the NVIDIA AI Enterprise (AIE) vGPU Manager image.
Procedure
Label the node with the name of the MIG-backed vGPU profile:
$ oc label node <node> --overwrite nvidia.com/vgpu.config=<profile>-
Replace
<node>with the fully qualified domain name (FQDN) of your compute node. Replace
<profile>with a supported MIG profile.Example command:
$ oc label node worker_1 --overwrite nvidia.com/vgpu.config=A30-1-6C
-
Replace