All Products
Search
Document Center

Container Compute Service:metrics-server

Last Updated:Mar 26, 2026

metrics-server is a resource monitoring component that Alibaba Cloud Container Service (ACS) builds on top of open-source Kubernetes Metrics Server. It collects resource usage metrics for all pods in your cluster and powers the Horizontal Pod Autoscaler (HPA). Call the Metrics API to retrieve these metrics programmatically.

Use cases

Use metrics-server for:

  • CPU/memory-based horizontal autoscaling — feed real-time resource usage to HPA to scale workloads automatically

  • Resource usage visibility — query current CPU and memory consumption per pod via the Metrics API

Do not use metrics-server when you need:

  • Historical or time-series metric data

  • High-accuracy resource usage for billing or auditing

  • Autoscaling based on custom or external metrics (use a custom metrics adapter instead)

How it works

metrics-server scrapes resource usage from each node's kubelet and exposes the aggregated data through the Kubernetes Metrics API. HPA queries this API to determine whether to scale workloads up or down.

metrics-server is installed by default in ACS clusters. No manual installation is required.

Enable CloudMonitor integration

To enable Kubernetes monitoring through CloudMonitor:

Prerequisites

Before you begin, ensure that you have:

  • Access to the ACS console

  • An existing ACS cluster

Steps

  1. Log on to the ACS console. In the left-side navigation pane, click Clusters.

  2. On the Clusters page, click the ID of the target cluster. In the left-side navigation pane, choose Operations > Add-ons.

  3. On the Logs and Monitoring tab, find metrics-server and click Configuration.

  4. In the dialog box, select Enables Kubernetes Monitoring of CloudMonitor and click OK.

Release notes

January 2024

Version

Release date

Changes

Impact

v0.3.9.7-85b3699-aliyun

January 2024

Improvements:

  • Optimized monitoring performance for large-scale clusters.

  • Stabilized HPA scaling during rolling updates. You must configure --enable-hpa-rolling-update-skipped=true to enable this behavior.

No impact on workloads.