Hello, I have set up a KVM cluster with Ceph as primary storage. I have several enterprise GPU (P4, P100 and V100) and consumer GPU (RTX3080). I would like to pass through the GPU to Windows and Linux VMs in an automated way. It would be ideal if I can use vGPU in enterprise GPUs, which would let user select (or fixed is also ok) vGPU profile (or any other way that can make GPU provisioning automatic). It would also be acceptable if I can create compute offerings that provides specific GPU model (entire GPU pass through). I have read that I can do this through API, but I didn’t find further documentation. If it’s currently not possible in KVM, will it be better if I switch to XCP-NG (and worth all the trouble)? Thank you for your time in answering my question.
Sincerely, Eleven
