If your Amazon EKS cluster runs Kubernetes 1.33 or older, you are already paying six times the normal control plane fee. AWS charges $0.10 per cluster per hour in standard support and $0.60 in extended support. Clusters on 1.31 lose extended support on 26 November 2026, and AWS will then upgrade them for you, with no warning and no way back. This guide shows where each EKS version stands today, what breaks between versions, and how to upgrade safely.
Where EKS Versions Stand in November 2026
AWS gives every Kubernetes version 14 months of standard support on EKS. After that, the version moves into 12 months of extended support at a higher price. That is 26 months in total. When extended support ends, AWS upgrades the control plane on its own.
Here is the current calendar, taken from the Amazon EKS version calendar:
| Version | EKS release | End of standard support | End of extended support | Status today |
|---|---|---|---|---|
| 1.37 | 1 Oct 2026 | 1 Dec 2027 | 1 Dec 2028 | Standard |
| 1.36 | 2 Jun 2026 | 2 Aug 2027 | 2 Aug 2028 | Standard |
| 1.35 | 27 Jan 2026 | 27 Mar 2027 | 27 Mar 2028 | Standard |
| 1.34 | 2 Oct 2025 | 2 Dec 2026 | 2 Dec 2027 | Standard, ends in weeks |
| 1.33 | 29 May 2025 | 29 Jul 2026 | 29 Jul 2027 | Extended ($0.60/hr) |
| 1.32 | 23 Jan 2025 | 23 Mar 2026 | 23 Mar 2027 | Extended ($0.60/hr) |
| 1.31 | 26 Sep 2024 | 26 Nov 2025 | 26 Nov 2026 | Extended, forced upgrade soon |
Two dates matter most right now:
- 26 November 2026. Extended support for 1.31 ends. After this date AWS can upgrade any 1.31 control plane at any time.
- 2 December 2026. Standard support for 1.34 ends. From that day, 1.34 clusters start paying the extended support fee.
Every version from 1.30 and below is already past the end of extended support. If you still see one of those in your account, AWS has either upgraded it already or is about to.
What Extended Support Really Costs
The fee looks small per hour. It adds up fast once you count every cluster, every hour, every month. These prices come from the Amazon EKS pricing page:
| Support tier | Per cluster per hour | Per cluster per month (730 hours) | 3 clusters per year |
|---|---|---|---|
| Standard support | $0.10 | $73 | $2,628 |
| Extended support | $0.60 | $438 | $15,768 |
So a team with dev, staging and production clusters on an old version pays about $13,000 a year more for the same clusters. Nothing else changes except the version number.
Three facts catch teams out:
- Extended support is on by default. Every new and existing cluster has its upgrade policy set to extended unless you change it. There is no prompt before the higher fee starts.
- Billing starts at midnight UTC on the day standard support ends.
- The forced upgrade does not stop the fee. A 1.31 cluster is moved to 1.32, which is also in extended support. You keep paying $0.60 per hour until you reach 1.34 or newer.
You can turn extended support off. If you do, AWS upgrades the cluster when standard support ends instead of charging you. That gives you a smaller bill but an upgrade date you do not control. For a full breakdown of every EKS line item, see our guide to Amazon EKS pricing.
What Happens When AWS Forces the Upgrade
When a version reaches the end of extended support, AWS moves the control plane to the oldest version it still supports. AWS says it cannot give an exact time. It can happen any time after the end date, and you will not get a notice first.
Only the control plane moves. The rest of the cluster stays where it was:
- Managed node groups and self-managed nodes stay on the old version. You must upgrade them yourself.
- EKS add-ons such as the VPC CNI, CoreDNS and kube-proxy are not upgraded. Old add-ons on a new control plane are a common cause of outages.
- EKS Auto Mode nodes may update on their own.
- You cannot roll it back. A cluster that AWS upgraded at the end of extended support is not eligible for rollback.
This is why waiting for AWS is the riskiest choice. You get an upgrade at a time you did not pick, a cluster with mismatched parts, and no undo button.
What Changed Between 1.31 and 1.37
Most of the pain in an upgrade comes from what was removed, not what was added. You can only move one minor version at a time, so a 1.31 cluster has to pass through every row below to reach 1.34. Review each change before you start. The details come from AWS release notes for extended and standard versions.
| Upgrading to | What to check first |
|---|---|
| 1.32 | The flowcontrol.apiserver.k8s.io/v1beta3 API is removed. Anonymous access now works only on health check endpoints. This is the last version with Amazon Linux 2 (AL2) AMIs. |
| 1.33 | No AL2 AMIs. Nodes must move to Amazon Linux 2023 or Bottlerocket. The Endpoints API is deprecated, so move scripts to EndpointSlices. |
| 1.34 | Nodes ship with containerd 2.1. AppArmor is deprecated. VolumeAttributesClass moves to the stable API. |
| 1.35 | The kubelet refuses to start on cgroup v1 by default. This is the last version that supports containerd 1.x. The --pod-infra-container-image kubelet flag is removed. IPVS mode in kube-proxy is deprecated. |
| 1.36 | gitRepo volumes no longer run. IP addresses and CIDRs with leading zeros or loose formats are rejected on create and update. Service externalIPs is deprecated. |
| 1.37 | SELinux mount labelling is on by default, which can stall Pods that share a volume. Auto Mode node pools default to "Balanced" consolidation. Some kubelet metrics and cAdvisor flags are removed. |
One more change sits outside the version table. The upstream Kubernetes project retired Ingress NGINX in March 2026. It gets no more security fixes. If you use it, plan a move to the Gateway API or another ingress controller as part of your upgrade work.
How to Upgrade an EKS Cluster Safely
An EKS upgrade has three parts: the control plane, the add-ons and the nodes. AWS runs the control plane upgrade when you ask for it. You own everything else.
Step 1: See what you are running
Start with an inventory. List every cluster, its version and its support status:
# Versions AWS offers today, with support end dates
aws eks describe-cluster-versions
# The version of one cluster
aws eks describe-cluster --name <cluster> --query cluster.version --output text
Do this in every region and every account. Old clusters often hide in a forgotten dev account.
Step 2: Read the upgrade insights
EKS checks your cluster for problems before you upgrade. Upgrade insights flag deprecated API calls, add-on version gaps and node version skew. Fix every item marked as an error before you go further.
aws eks list-insights --cluster-name <cluster> --region <region>
Step 3: Check your nodes and add-ons
- Move off AL2. From 1.33 onward there are no AL2 AMIs. Switch node groups to Amazon Linux 2023 or Bottlerocket first.
- Check containerd and cgroups. Custom AMIs must run cgroup v2 and containerd 2 before you pass 1.35.
- Line up add-on versions. List the versions that work with your target release, then plan to update each one.
aws eks describe-addon-versions --addon-name coredns --kubernetes-version <target-version>
Step 4: Upgrade the control plane, one version at a time
EKS only moves the control plane up one minor version per upgrade. If you are on 1.31 and want 1.34, that is three upgrades in a row. Test each one in staging first.
# With eksctl
eksctl upgrade cluster --name <cluster> --region <region> --approve
# Or with the AWS CLI
aws eks update-cluster-version --name <cluster> --kubernetes-version <target-version>
Keep kubectl on your laptop and CI runners within one minor version of the cluster.
Step 5: Update the add-ons
Update the EKS-managed add-ons to versions that match the new control plane. Self-managed tools, such as cert-manager or a monitoring agent, follow their own upgrade steps. Check each one against its own docs.
aws eks update-addon --cluster-name <cluster> --addon-name <addon-name> --addon-version <new-version>
Step 6: Upgrade the nodes
Managed node groups replace nodes with a rolling update. Pods move off each old node before it is removed. Set Pod disruption budgets so your apps stay up while this happens.
eksctl upgrade nodegroup --cluster <cluster> --name <nodegroup> --region <region> --kubernetes-version <target-version>
kubectl get nodes
Kubernetes allows nodes to run up to three minor versions behind the control plane. That is a safety net, not a plan. Bring nodes up to the same version as soon as you can.
Step 7: Test, then watch
Run your smoke tests and watch error rates, Pod restarts and latency for at least a day. Check that ingress, DNS and storage all behave as before.
Rolling Back an EKS Upgrade
Older guides, including the first version of this one, say an EKS control plane can never go back. That is no longer true. AWS now lets you roll back an in-place upgrade, with limits:
- Within 7 days. After 7 days the option is gone.
- One version only. You can go from 1.34 back to 1.33, not to 1.32.
- Control plane only. Managed node groups, self-managed nodes and add-ons must be rolled back by you, and before the control plane.
- Not after a forced upgrade. Clusters that AWS upgraded at the end of extended support cannot roll back.
- Extended fees come back. Rolling back to a version in extended support restarts the $0.60 per hour fee.
The full steps are in the AWS rollback guide. Rollback is a seat belt. Still back up your manifests and Helm values before every upgrade, for example with Velero.
Upgrade on Your Own Schedule
The safest habit is a fixed upgrade rhythm. Kubernetes ships a new minor version about every four months, so plan for two or three upgrades a year per cluster. Teams that do this never pay the extended support fee and never face a forced upgrade.
A few habits help:
- Set a calendar reminder 60 days before each end of standard support date.
- Upgrade dev first, then staging, then production, a week or two apart.
- Keep clusters in code. Terraform or eksctl config makes every upgrade repeatable. Our Kubernetes upgrade strategy guide covers zero-downtime patterns in more depth.
- Review the bill every month. An extended support line item is the earliest sign a cluster has fallen behind.
Where Atmosly Fits
Atmosly provisions EKS clusters, node groups and add-ons from a catalog, with guardrails you define. It can also import the clusters you already run. Add-on versions are pinned, so every change is deliberate, logged and easy to undo. The upgrade steps in this guide still apply. Atmosly keeps the clusters you upgrade consistent with each other.
Get Off Extended Support Before 26 November
If you have clusters on 1.31, 1.32 or 1.33, you are paying extra today, and 1.31 clusters face a forced upgrade this month. SquareOps is an AWS Advanced Consulting Partner with the EKS Service Delivery designation. Our engineers plan and run EKS upgrades, from the version audit through to node and add-on updates, with a tested rollback plan at every step. Talk to our Kubernetes team about getting your clusters back to standard support.