Skip to main content
The NVIDIA GPU Operator manages the GPU software stack on Kubernetes - driver, container toolkit, device plugin, DCGM, GPU Feature Discovery, and MIG configuration.

​
What Radar Shows

ClusterPolicy Detail View:
  • Operator state at a glance: ready, not ready (with an alert banner), or disabled
  • Enabled components as badges - Driver, Container Toolkit, Device Plugin, DCGM Exporter, GPU Feature Discovery, MIG Manager, and the rest - with disabled components shown muted
  • MIG strategy and operator namespace
NVIDIA’s ClusterPolicy shares its kind name with Kyverno’s ClusterPolicy. Radar disambiguates by API group everywhere - detail views, status badges, table columns, and filters all route to the right renderer.
NVIDIADriver Detail View:
  • Driver rollout state, with an alert when the driver DaemonSet hasn’t reached ready on all selected nodes
  • Driver type, version, image and repository, precompiled flag
  • Node selector for targeted driver deployments
Resource Browser: ClusterPolicy rows show state, enabled components, and MIG strategy; NVIDIADriver rows show state, type, and version.

​
Supported CRDs

CRDGroupTopologyDetail ViewAI Summary
ClusterPolicynvidia.com/v1-YesYes
NVIDIADrivernvidia.com/v1alpha1-YesYes

​
RBAC

Read access to the nvidia.com CRDs is granted by default (rbac.crdGroups.nvidia: true), so the operator’s resources render with no extra setup. Disable it if your org restricts Radar’s RBAC and you don’t run the GPU Operator.