beginnerOperations ~15 min

Diagnose a Pending GPU pod

Find and fix why a GPU pod won't schedule.

Scenario

A pod requesting a GPU is stuck Pending.

Prerequisites
  • ·A cluster with a Pending GPU pod (or create one requesting too many GPUs)
Tasks
0 / 3 tasks
  1. bash
    kubectl describe pod <pod> | sed -n '/Events/,$p'

    Expected · A reason like 'Insufficient nvidia.com/gpu' or a taint mismatch.

  2. bash
    kubectl get nodes -o custom-columns=NAME:.metadata.name,GPU:.status.allocatable.'nvidia\.com/gpu'
Success criteria
  • ✓The pod transitions to Running on a GPU node
Hints
  • ·No allocatable GPUs anywhere? It's an enablement problem — see the detection workflow.
  • ·Use the Troubleshooting decision tree for a guided diagnosis.