This project is archived and no longer maintained. Metrics Server was never meant to be used for non-autoscaling purposes and this project was more like a proof of concept back in the time I was working with k8s < 1.18. Since k8s 1.18+ kubelets expose more accurate resource usage metrics on /metrics/resource endpoint, you can scrape this endpoint instead.

Export metrics-server metrics to prometheus.

Metrics Server collects resource metrics from Kubelets and exposes them in Kubernetes apiserver through Metrics API for use by Horizontal Pod Autoscaler and Vertical Pod Autoscaler.

But why not use these metrics to create alerts with alertmanager or to visualize resource usage with grafana? Metrics-server-exporter aims to fill the gap between monitoring/alerting tools and metrics-server by exporting the data from the Metrics API to prometheus.

metrics_server_exporter_node_resource_capacity{node="node1",resource="cpu"} 42000    # millicores
metrics_server_exporter_node_resource_capacity{node="node1",resource="memory"} 42000 # kilobyte
metrics_server_exporter_node_resource_usage{node="node1",resource="cpu"} 1000        # millicores
metrics_server_exporter_node_resource_usage{node="node1",resource="memory"} 1000     # kilobyte
metrics_server_exporter_pod_resource_usage{namespace="kube-system",pod="metrics-server-exporter-68f7c886bf-wj8h6",resource="cpu"} 42 # millicores
metrics_server_exporter_pod_resource_usage{namespace="kube-system",pod="metrics-server-exporter-68f7c886bf-wj8h6",resource="memory"} 42 # kilobyte

• metrics-server must be deployed and fully operational in your cluster.
• prometheus to scrape the metrics-server-exporter metrics

Simply apply the deployment.yaml from this repo. This will create a ClusterRole, a ClusterRoleBinding, a ServiceAccount and the Deployment for the metrics-server-exporter.