beginnerOperations ~20 min

Configure DCGM GPU monitoring

Scrape GPU metrics with DCGM-exporter and view them.

Scenario

You need visibility into GPU utilization, temperature, and errors.

Prerequisites
  • ·Prometheus (kube-prometheus-stack)
  • ·GPU Operator (ships DCGM-exporter)
Tasks
0 / 3 tasks
  1. bash
    kubectl -n gpu-operator get pods -l app=nvidia-dcgm-exporter
  3. bash
    DCGM_FI_DEV_GPU_UTIL

    Expected · A time series per GPU.

Success criteria
  • ✓DCGM metrics appear in Prometheus
  • ✓GPU utilization/temperature/power are visible
Hints
  • ·See the Observability section for the full metric bands and alerts.