Node Autoscaling with Karpenter — Part 2

In the previous chapter, you got to know how Karpenter works in much more detail, and the most important aspects when using Karpenter. One in particular is that Karpenter works in tandem with kube-scheduler. We also explored how to define flexible, well-scoped NodePools and how to configure your workloads so that Karpenter can provision the right capacity based on application needs. But launching the right nodes is only part of the story. You also need to adapt to changes over time, reduce inefficiencies, and maintain alignment with the evolving state of your cluster and infrastructure policies.

In this chapter, we’ll shift our focus from provisioning toward disruption and termination controllers. We’ll explore how Karpenter replaces nodes that are no longer needed, and the handling of spot interruptions or node expirations. You’ll learn that Karpenter not only does an efficient job when launching nodes but also...