Hi,
I'm working with a client on getting ArcGIS Enterprise on Kubernetes deployed on their Nvidia DGX H100. We're in the pre-install (preparation) stage.
Part 1: Has anyone managed to deploy ArcGIS Enterprise on Kubernetes on a Nvidia DGX H100 system? 😅
Part 2: In the documentation "Configure GPU-enabled nodes", it indicates,
Verify that your instance has the NVIDIA device plug-in for Kubernetes installed.
If the system has Nvidia GPUs but does *not* use the Nvidia device plug-in for Kubernetes, will I be able to deploy with GPU support? They currently use the "HAMi" device plugin.
To my knowledge, the device plugin operates inside kubelet, and the pod only interacts with the kubelet via sockets. So the pod shouldn't care, right?
Technically (simplified) - if the pod asks kubelet for 'nvidia/gpu' and magically gets an Nvidia GPU,
if it looks/smells/works like Nvidia, it _shouldn't_ matter what device plugin was involved, right?
Thanks,
Derek
