Capacity to use multiple GPUs with gpu-operator

Hello !

I’m working on the capacity to deply on a local k8s cluster, a node equipped with 2 GPUs (to have a physical segregation for critical applications). We plan to use the gpu-operator to manage the configuration of the nodes. I just have a basic question, before spending two many times on this activity : is it theoretically possible to manage this configuration with the gpu-operator or for the while it only manages configuration of nodes equipped with a single GPU ?

Thanks !

Hello @sebastien.bonnet.eu and welcome to the NVIDIA developer forums.

If I understand it correctly, gpu-operator works transparently with respect to the number of GPUs available in each node. But I am not the expert on this and it would be a good idea to verify this by opening an issue on the corresponding Github page. That is usually the place where people supporting this tool answer questions.

Thanks!