Property

dcgm_exporter

NVIDIA DCGM Exporter exposes GPU telemetry as Prometheus metrics, typically on port 9400. Only static hardware information is collected; live telemetry such as utilization and temperature is not stored.

Property Name Type Description Required
driver_version string Version of the NVIDIA driver.
gpu_count integer Number of GPUs reported by the exporter.
gpus array of DcgmExporterGpu List of GPUs on the host.
hostname string Hostname of the GPU host as reported by the exporter.

DcgmExporterGpu

Property Name Type Description Required
index integer Index of the GPU on the host (the "gpu" label).
memory_total_mib integer Total framebuffer memory of the GPU in MiB.
model string Model name of the GPU, ex: NVIDIA H100 80GB HBM3
pci_bus_id string PCI bus ID of the GPU, ex: 00000000:18:00.0
uuid string Unique identifier of the physical GPU, ex: GPU-bd14493d-d38c-85b1-0000-000000000000


↑ TopSearch Queries
SearchNVIDIA DCGM Exporters
product:"DCGM Exporter"
SearchDCGM Exporters with NVIDIA H100 GPUs
product:"DCGM Exporter" "H100"

↑ TopExample
{
    "hostname": "gpu-node-01.example.com",
    "driver_version": "570.211.01",
    "gpu_count": 2,
    "gpus": [
        {
            "index": 0,
            "uuid": "GPU-00000000-0000-0000-0000-000000000000",
            "model": "NVIDIA H200",
            "memory_total_mib": 143771,
            "pci_bus_id": "00000000:18:00.0"
        },
        {
            "index": 1,
            "uuid": "GPU-00000000-0000-0000-0000-000000000001",
            "model": "NVIDIA H200",
            "memory_total_mib": 143771,
            "pci_bus_id": "00000000:29:00.0"
        }
    ]
}