Amazon EKS now has Kubernetes version rollbacks! Even if an upgrade goes wrong, you can revert within 7 days

Hi everyone, it's Shii! Today I found news about a new Amazon EKS feature, and I just can't stop feeling excited. I mean, "you can upgrade and still revert if push comes to shove" is such a reassuring thing for anyone running clusters, right? Let me tell you all about it!

AWS Blog aws.amazon.com

What was announced?

On July 1, 2026, AWS announced "Kubernetes version rollbacks," a feature that lets you safely revert the Kubernetes version of an Amazon EKS cluster. Until now, even if you found a problem after upgrading a minor version, you couldn't easily go back. But with this feature, you can revert to a previous version within 7 days of an upgrade.

You can only revert one minor version at a time. And the key point is that it isn't just a rewind, you return to a validated state that was actually running in production. For example, if a troublesome problem shows up after upgrading from Kubernetes 1.34 to 1.35, you can go back to 1.34 within 7 days.

Before you run a rollback, an automatic check via "cluster insights" runs to confirm node version compatibility and add-on dependencies. The safety-first design is a point I love about it.

The story so far

Actually, open-source Kubernetes itself doesn't come with a mechanism for control plane rollback to begin with. In other words, "once you upgrade, there's no going back" was the basic rule.

Because of this one-way nature, operations teams had no choice but to be extremely careful. They had to set bake periods, roll out gradually across cluster groups, build approval workflows, and generally assemble labor-intensive compensating measures with every upgrade.

Especially in heavily regulated environments, the lack of a reliable recovery path became a sticking point, and there were even cases where upgrades themselves got postponed. As a result, clusters kept running on old versions without security patches applied, while support deadlines crept ever closer, a genuinely tricky situation.

What changes

With this new feature, an upgrade shifts from a "one-shot with no going back" to a "do-over-friendly operation with a 7-day grace period." Even if unexpected issues show up on the new version, you can calmly revert to the previous version and give yourself time to investigate the cause.

I think this is also encouraging for teams that tended to put off upgrades. It might help you avoid missing the right timing to apply security patches, and the larger your clusters, the more peace of mind it should add.

Dive Deep

From here, let me dig properly into the technical details.

Scope and time window

• You can roll back within 7 days of an upgrade
• You can only revert one minor version at a time (stepwise reverts are required)
• It applies to EKS standard support and extended support versions

How the pre-checks work Before you run a rollback, cluster insights automatically checks node version compatibility and add-on dependencies. If there are potential issues, you can catch them here, which reduces the risk of suddenly reverting into a broken state. If you're in a hurry and want to skip the pre-check, it seems you can add the --force flag in the CLI like this to skip it.

aws eks update-cluster-version --name my-cluster --version 1.34 --force

(Check the actual documentation for the option name and syntax. I'd appreciate it if you took the command example as just an illustration.)

For EKS Auto Mode On clusters using EKS Auto Mode, you can roll back the control plane and managed nodes at the same time. And when it does, it respects Pod Disruption Budgets while it works. That said, this can make the rollback take longer. If you want it to finish faster, you have the option of manually adjusting or removing the disruption budgets. There's also a cancellation API available for when you want to stop a node rollback partway through.

About pricing There's no additional charge for the control plane rollback feature itself. You only pay the usual EKS usage fees and compute resource costs. It's a nice touch that the design makes it easy to try.

The flow (console) In the Amazon EKS console, you select the target cluster, check the rollback options from the settings page, review the cluster insights results before executing, and then run it. Since you can confirm what's being checked in advance, it should reduce the worry of suddenly impacting production.

Wrap-up

The "Kubernetes version rollbacks" feature added to Amazon EKS strikes me as pretty practical. To recap the key points:

• Within 7 days of an upgrade, you can revert one minor version back to a validated previous version
• Before you run it, cluster insights automatically performs compatibility checks
• With EKS Auto Mode, you can roll back the control plane and managed nodes together
• It's available at no additional cost

That was the gist. It's a quiet but welcome update that might make "upgrades are scary, so let's put it off" a little less common. Check out the original for the details.