Hi everyone,
We are considering using Kubernetes to manage batch jobs on an on-prem GPU
cluster.
However, I had a hard time looking for updated information about GPU support in
Kubernetes.
Here it said only 1 GPU per machine is supported:
https://github.com/kubernetes/kubernetes/pull/24836
Is
Awesome!
I checked out the branch, rebase with master, compiled it and deploy on a
single-node Ubuntu 14.04 machine.
However, when I run "kubectl describe nodes", it shows:
Capacity:
alpha.kubernetes.io/nvidia-gpu:0
Allocatable:
alpha.kubernetes.io/nvidia-gpu:0
I am pretty