BeaverDeck logo
BeaverDeck

Home / BeaverDeck / Docs / Insights Guide / GPU Insights / GPU Allocation Pressure

GPU Allocation Pressure

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 typegpu-allocation-pressure
Insights sectionGPU Insights
Alert severityWarning at 80%; critical at 95%

When It Reports A Finding

Active pod GPU requests reach at least 80% of allocatable GPUs on a GPU node or across the selected namespaces. The severity becomes critical at 95%.

Why This Is A Problem

Little unallocated capacity remains for new workloads, failover, rolling updates, or autoscaling. Pending GPU pods become more likely as allocation approaches capacity.

Recommended Response

  1. Review requested versus allocatable GPUs at node and selected-namespace scope.
  2. Remove stale workloads and right-size GPU requests where the workload can use smaller allocations.
  3. Redistribute workloads or add compatible GPU nodes before planned demand or maintenance.

Scope And Limitations

This is scheduling allocation pressure, not actual GPU utilization. A requested GPU can be idle, while a busy GPU still counts as one allocation.

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.