nvidia-license-server-exporter

A NVIDIA Cloud License Service (CLS) exporter with:

• Prometheus pull (/metrics)
• Optional OTEL metrics push (OTLP gRPC)

The exporter is intentionally scoped to CLS only (no DLS support).

Required credentials and IDs

• API key with Licensing State access
• Org ID/name (for example: lic-...)

Optional:

• Service instance ID (x-nv-service-instance-id) for environments that require it

Configuration

CLS and server

• NVIDIA_API_KEY (required)
• NVIDIA_ORG_NAME (required)
• NVIDIA_API_BASE_URL (optional, default https://api.licensing.nvidia.com)
• NVIDIA_SERVICE_INSTANCE_ID (optional)
• LISTEN_ADDRESS (optional, default :9844)
• METRICS_PATH (optional, default /metrics)
• SCRAPE_TIMEOUT (optional, default 20s)
• CACHE_TTL (optional, default 60s)
• PARALLELISM (optional, default 8)

If LISTEN_ADDRESS is unset and PORT is set (for example on Railway), the exporter listens on :$PORT.

OTEL push (optional)

• OTEL_ENABLED (optional, default false)
• OTEL_ENDPOINT (optional, default 127.0.0.1:4317)
• OTEL_SERVICE_NAME (optional, default nvidia-license-server-exporter)
• OTEL_SERVICE_INSTANCE_ID (optional, default hostname)
• OTEL_INSECURE (optional, default true)
• OTEL_PUSH_INTERVAL (optional, default 60s)

Flags are also available in -kebab-case (for example -otel-enabled, -otel-endpoint).

Run

go mod tidy
go run ./cmd/nvidia-license-server-exporter \
  -nvidia-api-key "$NVIDIA_API_KEY" \
  -nvidia-org-name "$NVIDIA_ORG_NAME"

Endpoints:

• GET /metrics
• GET /healthz

Shared cache behavior

Prometheus pull and OTEL push use the same snapshot cache.

• If cache is fresh (CACHE_TTL), no CLS API call is made.
• If cache is stale, one refresh call updates cache for both pull and push.
• If refresh fails and a stale snapshot exists, stale data is still emitted with nvidia_cls_up=0.

Recommended default:

• CACHE_TTL=60s
• OTEL_PUSH_INTERVAL=60s

Exported metrics

Core health:

• nvidia_cls_up
• nvidia_cls_scrape_duration_seconds
• nvidia_cls_scrape_timestamp_seconds

Entitlement:

• nvidia_cls_entitlement_total_quantity

Server:

• nvidia_cls_license_server_info
• nvidia_cls_license_server_feature_total_quantity
• nvidia_cls_license_server_feature_active_leases

All metrics include constant label org_name="<your org id>".

Prometheus scrape config example

scrape_configs:
  - job_name: nvidia_cls
    static_configs:
      - targets:
          - localhost:9844