From Heapster to Metrics Server #

The critical element in scaling Pods is the Kubernetes Metrics Server. You might consider yourself a Kubernetes ninja and yet never heard of the Metrics Server. Don’t be ashamed if that’s the case. You’re not the only one.

If you started observing Kubernetes metrics, you might have used Heapster. It’s been around for a long time, and you likely have it running in your cluster, even if you don’t know what it is. Both the Metrics server and Heapster serve the same purpose, with one being deprecated for a while, so let’s clarify things a bit.

Early on, Kubernetes introduced Heapster as a tool that enables Container Cluster Monitoring and Performance Analysis for Kubernetes. It’s been around since Kubernetes version 1.0.6. You can say that Heapster has been part of Kubernetes’ life since its toddler age. It collects and interprets various metrics like resource usage, events, and so on. Heapster has been an integral part of Kubernetes and enabled it to schedule Pods appropriately. Without it, Kubernetes would be blind. It would not know which node has available memory, which Pod is using too much CPU, and so on. But, just as with most other tools that become available early, its design was a “failed experiment”. As Kubernetes continued growing, we (the community around Kubernetes) ...