AWS just gave Kubernetes users on its EKS platform something a lot of them have been building themselves for years out of necessity: direct control over specific settings inside the Kubernetes control plane, without having to run and maintain custom components to get there.
The new configuration options touch four main areas: how the scheduler places pods across nodes, how quickly autoscaling responds to demand, how long Kubernetes retains event data, and which ports NodePort services can use. None of that sounds thrilling on its own, but for teams running high-churn workloads, batch processing, AI compute, CI/CD pipelines that spin containers up and down constantly, these are exactly the settings that used to require a bespoke solution nobody particularly wanted to maintain.
Take pod placement, for example. Administrators can now switch the scheduler between spreading workloads across nodes with more available capacity or packing them onto nodes that already run heavier loads. That second option sounds counterintuitive at first, but it actually helps consolidate workloads onto fewer machines, which means fewer instances sitting around partially idle and, ideally, a smaller compute bill.
Autoscaling gets a similar upgrade. Teams can now adjust how often the Horizontal Pod Autoscaler checks metrics and makes scaling decisions. Shorten that window, and workloads respond to demand spikes faster, though obviously at the cost of more frequent evaluation overhead.
Event retention is where things get genuinely practical. Kubernetes normally keeps events around for an hour by default, but workloads that constantly create and destroy pods can flood storage and slow down API server operations as a result. AWS now lets customers set retention anywhere between 10 and 60 minutes, trading historical debugging data for cleaner, faster performance, depending on what actually matters more for a given workload.
NodePort configuration rounds out the update, letting teams adjust which port ranges Kubernetes uses for external traffic. That flexibility matters most for organizations migrating older applications that expect specific ports, since it avoids the usual workaround of adding a proxy layer just to make things fit.
The feature is available now on EKS clusters running Kubernetes 1.31 or later, across AWS commercial regions, GovCloud, and China regions where EKS operates. It’s free to use, though a couple of the more advanced parameters, including the configurable HPA sync period, require the separately billed EKS Provisioned Control Plane tier.
