Skip to content

GPU Devices

Purpose

GPU / PCI passthrough inventory: host VGA/3D devices, allocation to VMs, plus a vGPU orchestration panel for mdev profiles and assign actions.

When to use it

  • See which host GPUs are free vs allocated
  • Confirm IOMMU group, driver, and PCI address before passthrough
  • Review VM GPU assignment rows
  • Assign a GPU resource or host device to a VM from the vGPU panel

How to get there

  • Route: /gpu
  • Nav: Cluster → GPU Devices
  • Related: /device-plugins, /vms, /nodes

Operate from the console (UX)

  1. Open /gpu from the sidebar, workspace switcher, or command palette (⌘K / Ctrl+K).
  2. Read the PageHead / deck line — confirm it matches: GPU / PCI passthrough inventory: host VGA/3D devices, allocation to VMs, plus a vGPU orchestration panel for mdev profil….
  3. Use namespace / label filters; open an object for inspector details.
  4. Mutations (drain, resize PVC, attach NAD) go through confirm sheets — wait for status refresh.
  5. If empty: broaden namespace filter, confirm RBAC, and that the cluster API is reachable.
  6. Empty / fail: No data → check auth, namespace filter, and that required operators (KubeVirt/CDI/monitoring) are installed; purpose was: GPU / PCI passthrough inventory: host VGA/3D devices, allocation to VMs, plus a vGPU orchestration p.
  7. Success: Live inventory or job progress matches the deck line; you can navigate to a related center without re-auth.

Use https://<host>:30050/ (NodePort) or your ingress URL — never publish lab IPs. HTTPS is required for in-browser VNC. Browse lists are safe; power actions live on workbenches.