furiosa-metrics-exporter

Overview

The image runs a single binary that periodically collects metrics from the NPUs on the node and serves them at the /metrics HTTP endpoint for Prometheus to scrape. It must run on nodes that have FuriosaAI NPUs and needs access to the host's NPU devices and driver. The container starts the binary automatically through its default command.

Configuration

Behavior is controlled by command-line flags:

•  --port — port number used by the metrics server (required).
• --interval — collection interval in seconds (required).
• --node-name — name of the node the exporter is running on (optional).
• --kube-resources-label — enable injection of Kubernetes resource labels such as namespace, pod, and container (optional).

Deploying with Helm

The recommended way to deploy on Kubernetes is with the official Helm chart, published at https://github.com/furiosa-ai/helm-charts. The chart also creates a Service with Prometheus scrape annotations so metrics are picked up automatically. Options can be adjusted in charts/furiosa-metrics-exporter/values.yaml before installing.

helm repo add furiosa https://furiosa-ai.github.io/helm-charts
helm repo update
helm install furiosa-metrics-exporter furiosa/furiosa-metrics-exporter -n kube-system

Verifying

After deployment, confirm the exporter is serving metrics:

kubectl port-forward <metrics-exporter-pod> 9090:<port> -n kube-system
curl http://localhost:9090/metrics | grep furiosa_npu

You should see the furiosa_npu_* metrics for each NPU on the node.