BeaverDeck logo
BeaverDeck

Home / BeaverDeck / Docs / Insights Guide / GPU Insights / GPU Node Workload Mix

GPU Node Workload Mix

BeaverDeck uses this check to identify a specific gpu condition that may need operator review.

Permissions: viewing checks requires insights: view. Opening a linked object or logs requires the corresponding resource permission, and the BeaverDeck ServiceAccount must be allowed to read the Kubernetes resources used by the check. Suppressing a finding requires insights: edit and affects all users.
Check typenon-gpu-pods-on-gpu-node
Insights sectionGPU Insights
Alert severityWarning

When It Reports A Finding

An active, non-DaemonSet Pod without a GPU request is scheduled on a GPU node.

Why This Is A Problem

CPU-only workloads can consume CPU, memory, and pod slots on scarce GPU nodes, reducing room for workloads that actually require the accelerator.

Recommended Response

  1. Confirm whether each listed non-GPU pod is intentionally placed on the GPU node.
  2. Use GPU-node taints and explicit tolerations, node affinity, or selectors to reserve the pool where appropriate.
  3. Move general workloads to non-GPU pools while preserving required node agents.

Scope And Limitations

DaemonSet-managed pods are intentionally ignored. Some data, sidecar, or control workloads may legitimately need the same node despite not requesting a GPU.

After remediation: refresh GPU Insights and verify the underlying resource or metric. Suppress the finding only when the condition is intentional and its risk is accepted.