Scaling GPU Worker Nodes

NVIDIA AI Enterprise 2.0 or later

Scaling of GPU worker nodes can be done via combining the use of the OpenShift Console and VMware vCenter. In this example we will have deployed a cluster via the IPI method and already attached a virtual GPU to our OpenShift worker VMs.

openshift-appendix1.png

To scale navigate to Compute and MachineSets.

Select the MachineSet for your GPU Accelerated cluster

openshift-appendix2.png

Select the number of machines under Desired Count

openshift-appendix3.png

Increase the number of machines for your cluster

openshift-appendix4.png

Wait until the Desired, Current, Ready, and Available, counts are all equal

openshift-appendix5.png

Navigate to VMware vCenter and select the newly created worker VM in you inventory.

For each new worker VM perform the follow steps:

Power down the VM and edit the settings

openshift-appendix6.png

Add a new device and click PCI Device

openshift-appendix7.png

Attach the appropriate vGPU profile or pass-through GPU

openshift-appendix8.png

Power on the VM.

© Copyright 2022-2023, NVIDIA. Last updated on Jan 9, 2023.