deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

EKS clusters on Kubernetes 1.34 face sixfold control plane price rise from 2 December

EKS clusters on Kubernetes 1.34 leave standard support on 2 December 2026, when the control plane rate rises from $0.10 to $0.60 per hour. A dev.to guide explains how to audit clusters and upgrade first.

EKS clusters on Kubernetes 1.34 face sixfold control plane price rise from 2 December

EKS clusters still running Kubernetes 1.34 fall out of standard support on 2 December 2026, after which the control plane is billed at $0.60 per hour instead of $0.10 — roughly $438 a month per cluster rather than about $73. A guide published on dev.to on 19 November walks through how to inventory clusters, stop the higher rate from starting quietly, and run upgrades that repeat every few months.

How the two support phases are priced

According to the post, which cites AWS pricing, each Kubernetes version on EKS receives 14 months of standard support from its release on the service, with the control plane at $0.10 per cluster hour. Extended support then runs for 12 more months at $0.60 per cluster hour. Once both windows close, 26 months after release, AWS upgrades the cluster itself.

The charge is per cluster, not per node. Nodes, Fargate pods, load balancers and EBS volumes cost the same in either phase, and only the control plane line item changes — which the author argues makes the jump easy to overlook in Cost Explorer. A team with production, staging and dev clusters parked on an old version pays about $1,095 a month more than needed, close to $13,000 a year.

The lifecycle calendar

Version Standard support ends Extended support ends
1.32 23 March 2026 (already extended) 23 March 2027
1.33 29 July 2026 (already extended) 29 July 2027
1.34 2 December 2026 2 December 2027
1.35 27 March 2027 27 March 2028
1.36 2 August 2027 2 August 2028
1.37 1 December 2027 1 December 2028

Clusters on 1.32 or 1.33 are already billing at the extended rate, and 1.34 has only weeks of standard support left. The post names 1.35 and 1.36 as the safest upgrade targets, since both stay in standard support well into 2027.

Setting the upgrade policy

Every EKS cluster carries an upgrade policy, and the default lets it drift into extended support and the higher price. Setting it to STANDARD, for example aws eks update-cluster-config --name staging --upgrade-policy supportType=STANDARD, makes AWS upgrade the cluster to the next version when standard support ends instead of billing $0.60.

Two caveats apply. The setting only works while the cluster is still on standard support, so a cluster already paying the extended rate must be upgraded to return to $0.10. And the automatic upgrade happens on AWS's timing, not the operator's — the author frames it as a backstop against the charge, not a replacement for a tested upgrade plan.

Terraform exposes the same control through the upgrade_policy block on the aws_eks_cluster resource. It arrived with the 5.x series of the AWS provider, and a plan should show an in-place update rather than cluster replacement.

What to check before upgrading

The post attributes most failed EKS upgrades to components still calling APIs that the next version removed. Upgrade insights, via aws eks list-insights, flag these ahead of time, and anything not marked PASSING needs fixing first.

Next, in order: EKS add-ons — VPC CNI, CoreDNS, kube-proxy and the EBS CSI driver — whose supported versions are tied to the Kubernetes release; self-managed Helm charts such as ingress controllers, cert-manager, the AWS Load Balancer Controller and Karpenter; and node AMIs, where Amazon Linux 2 images stop at 1.32, forcing a move to Amazon Linux 2023 or Bottlerocket for anything newer.

One minor version per round

EKS raises the control plane by a single minor version per upgrade, so a 1.33 cluster needs two rounds to reach 1.35. Each round upgrades the control plane first (roughly 10 to 20 minutes, with the API still available), then add-ons to matching versions, then managed node groups through a rolling node replacement. PodDisruptionBudgets that permit zero disruptions will stall that rollout until it times out, so they deserve a check beforehand. The author's closing habits: upgrade staging first, let it soak under normal traffic, and write down what was done so the next cycle takes half the time.

Why it matters

The extended-support premium buys nothing new: the cluster is identical, and the money only keeps an aging version patched. The fee is charged per cluster, so it scales with fleet size while hiding in a single Cost Explorer line. With 1.33 already on the extended rate and 1.34 crossing over on 2 December, the immediate work is an inventory of every cluster's region, version and upgrade policy, followed either by planned upgrades to 1.35 or 1.36 — or, at minimum, an upgrade policy that guarantees the sixfold rate can never begin unnoticed.

  • #aws
  • #kubernetes
  • #amazon-eks
  • #cloud-costs
  • #upgrades

Related posts