Skip to content
UMKIMDocs
Search limited to 0.1.0. Switch to the latest version for up-to-date results.

GPU host prep

This content is for 0.1.0. Switch to the latest version for up-to-date documentation.

GPU workloads require preparation outside the FCJ bootstrap lifecycle:

  1. Install and validate a compatible host driver on every intended node.
  2. Configure the container runtime and Kubernetes device plugin after the cluster reaches Ready.
  3. Confirm allocatable GPU capacity before enabling GPU-targeted placement.
  4. Install the kubevirt provider only when the host and cluster requirements have been independently validated.

Driver compatibility, hardware support, capacity discovery, and workload reliability are operator responsibilities. GPU and LLM paths are experimental in v0.1.0 and carry no production support claim.