NVIDIA DCGM Exporter
A Prometheus exporter for collecting NVIDIA GPU telemetry from supported hosts and clusters.
Published by NVIDIA
How to use it
Monitor GPU utilization, memory, temperature, and available error counters alongside application throughput. Align timestamps when comparing a latency spike with a GPU resource change.
Run the exporter on a supported host and scrape its metrics with Prometheus. Confirm the GPU driver and DCGM requirements for the deployment before scheduling alerts.
- Compatibility
- NVIDIA GPUs, Prometheus, Docker, Kubernetes
- Monitoring areas
- GPU telemetry, Training infrastructure metrics, Prometheus export
- Access
- Local exporter. Requires supported NVIDIA hardware and drivers.