-
Epic
-
Resolution: Unresolved
-
Major
-
None
-
Support GPU workloads on HCP KubeVirt platform
-
0.42
-
False
-
-
False
-
- We have the ability to assign GPUS to kubevirt NodePools
- We have a CI lane to verify gpu assignment to kubevirt nodepools
-
Yellow
-
To Do
-
OCPSTRAT-1043 - [Tech Preview] HCP KubeVirt Nvidia GPU support
-
OCPSTRAT-1043[Tech Preview] HCP KubeVirt Nvidia GPU support
-
27% To Do, 9% In Progress, 64% Done
-
dev-ready, doc-ready, po-ready, qe-ready, ux-ready
-
-
---
-
---
GOAL:
Support running workloads within HCP KubeVirt clusters which need access to GPUs.
Accomplishing this involves multiple efforts
- The NodePool API must be expanded to allow assignment of GPUs to the KubeVirt worker node VMs.
- ensure nvidia operator works within the HCP cluster for gpus passed through to KubeVirt VMs
- Develop a CI environment which allows us to exercise gpu passthrough.
Diagram of multiple nvidia operator layers
https://docs.google.com/document/d/1HwXVL_r9tUUwqDct8pl7Zz4bhSRBidwvWX54xqXaBwk/edit