To offer GPUs in VMs in OpenShift virtualization we need to figure out how resourcequotas's will apply when the GPUs are configured in PCI passthrough mode. This also ties in on how we will manage said quotas from coldfront and ultimately generate invoices for these.
Completion Criteria
We can enforce quotas on passthroughed GPUs.
Description
When a GPU is configured in passthrough mode they show up as new allocatable resources. For instance, nvidia.com/A100_SXM4_40GB appears as a resource for A100 GPUs, and nvidia.com/GV100GL_Tesla_V100 for V100 GPUs. As a result, we will need to update the resource quotas for projects and adjust Coldfront configuration to manage access to these resources.
Motivation
To offer GPUs in VMs in OpenShift virtualization we need to figure out how resourcequotas's will apply when the GPUs are configured in PCI passthrough mode. This also ties in on how we will manage said quotas from coldfront and ultimately generate invoices for these.
Completion Criteria
We can enforce quotas on passthroughed GPUs.
Description
When a GPU is configured in passthrough mode they show up as new allocatable resources. For instance, nvidia.com/A100_SXM4_40GB appears as a resource for A100 GPUs, and nvidia.com/GV100GL_Tesla_V100 for V100 GPUs. As a result, we will need to update the resource quotas for projects and adjust Coldfront configuration to manage access to these resources.
Original issue where this is discussed: https://github.com/nerc-project/operations/issues/725#issuecomment-2474796687
Completion dates
Desired - 20YY-MM-DD Required - TBD