Amazon EKS Now Supports Certificate Authority (CA) Rotation with Automated Lifecycle Management
Amazon EKS's new CA rotation lets you rotate a cluster's certificate authority without downtime, through a managed lifecycle with automated safeguards. This post covers how it works, how to tell when your own clusters need to act, and how to carry out the rotation step by step.
Introduction
On August 20, 2026, AWS announced certificate authority (CA) rotation for Amazon EKS. The headline feature of this update is that you can now rotate a cluster's CA before it expires, through a managed lifecycle with automated safeguards.
If you've been running an EKS cluster since around 2018 or 2019 without giving much thought to its CA, this update is worth a closer look. It exists specifically to address the CA expiration issue that clusters from that era are now approaching.
This post covers what changed, how CA rotation works under the hood, and how to assess when and how your own clusters should act. It's written for infrastructure and platform engineers running EKS, as well as SREs responsible for certificate lifecycle management.
What Changed
Every Amazon EKS cluster has its own CA that enables encrypted connections to the cluster's Kubernetes API. Clusters created since EKS launched in 2018 have CAs with a 10-year validity period, and clusters from that era are now approaching the point where CA rotation activities should begin.
Aspect | Before | After CA rotation |
|---|---|---|
Rotation method | No standardized mechanism; manual and complex | Managed lifecycle, available via the AWS Console, CLI, and API |
Managing expiration risk | Customers had to track and manage this individually | Automated safeguards, including advance notifications before expiration |
Handling a failed migration | No clear rollback path | A rollback capability to revert to the previous CA |
The Previous Challenge
Until now, EKS had no standardized mechanism for safely rotating a cluster's CA. Tracking the CA's expiration and deciding when to act was left entirely up to the customer. This was a particular risk for organizations that created clusters around 2018 and have run them continuously since, where the CA expiration risk could accumulate without much visibility. Because an expired CA breaks connectivity to the Kubernetes API server, this is fundamentally an availability problem for the cluster.
How CA Rotation Solves This
With this update, EKS manages the CA rotation lifecycle and clarifies how responsibility is divided between AWS and the customer.
Responsibility | Scope |
|---|---|
AWS is responsible for | Managing the rotation lifecycle, automatically updating AWS-managed components to trust the successor CA, and automatically updating EKS Auto Mode instances and AWS Fargate nodes |
You are responsible for | Replacing your worker nodes and updating external clients (kubeconfig files, CI/CD pipelines, monitoring tools, and so on) to trust the successor CA |
Building on this division of responsibility, EKS provides the following safeguards. You're notified in advance as the CA's expiration approaches, and a successor CA is automatically appended if you haven't created one yourself. If you don't activate the successor CA on your own schedule, EKS automatically activates it for you. And if an issue comes up during the transition to the successor CA, a rollback capability lets you revert to the previous CA.
CA rotation is available at no additional cost in all commercial AWS Regions.
How CA Rotation Works
CA rotation doesn't switch CAs instantly. It's a multi-stage process — appending a successor CA, distributing it, and activating it — during which the outgoing and successor CAs are trusted simultaneously. This is what makes it possible to migrate incrementally without taking the cluster down.
Stage 1: Append a Successor CA
In the first stage, a successor CA is appended to the cluster. From this point, the cluster trusts both the CA it has been using (the outgoing CA) and the newly appended successor CA. Certificates continue to be issued by the outgoing CA, so nothing breaks yet.
As long as the cluster is in an active state, you can append a successor CA yourself, on your own timeline, using the AWS CLI, EKS API, Console, or infrastructure as code such as AWS CloudFormation. If you don't, AWS appends one automatically. This stage is also a good time to start identifying the worker nodes and external clients that will need to be updated.
Stage 2: Distribute the Successor CA
Once the successor CA is appended, AWS automatically updates AWS-managed components — the control plane, EKS Auto Mode instances, and AWS Fargate nodes — to trust it. You can track this progress through the CA's distribution status, and the successor CA can't be activated until distribution is complete.
After distribution is complete, it's your turn: you need to update the worker nodes you manage and your external clients to trust the successor CA.
Stage 3: Activate the Successor CA
Once distribution to all AWS-managed components is complete, the successor CA becomes eligible for activation. We recommend activating it on your own timeline, after confirming that the worker nodes you manage and your external clients have been updated. If you take no action, AWS automatically activates the successor CA as the expiration deadline approaches.
After activation, the cluster issues all new certificates signed by the successor CA. The outgoing CA remains trusted but is no longer used for signing. A rollback window is available for a limited period after activation, letting you revert to the outgoing CA if a problem comes up.
The Dual Trust Period
The time between appending a successor CA and retiring the outgoing CA is called the dual trust period. During this window, the cluster trusts both CAs simultaneously, so you don't need to switch every worker node and external client at once — you can update them incrementally. This is what allows CA rotation to proceed without disruption.

How Activation Affects Connectivity
When a client connects to the API server, it verifies that the certificate the server presents was signed by a CA the client trusts. Once the successor CA is activated, the API server starts presenting certificates signed by that successor CA.
Clients that have already been updated to trust the successor CA connect without any issue. Clients that haven't been updated can't verify the certificate's signer, so their connection fails. This is exactly why worker nodes and external clients need to be updated to trust the successor CA before activation.
One more detail worth knowing: the CA created when a cluster is first created has a 10-year validity period, but successor CAs created through rotation have a 5-year validity period. In other words, once a cluster has been through its first rotation, it will repeat this cycle roughly every five years.
Is Your Cluster Affected?
CA rotation is something every EKS cluster will eventually go through. This section looks at two questions: when should you start thinking about it, and should you let AWS's automation handle it or take the lead yourself?
When to Get Started
You can check when your cluster's CA expires using
describe-certificate-authority . If your cluster was created around 2018–2019, 2028–2029 is a reasonable ballpark for when you'll need to act.Even if you don't take any action yourself, AWS provides notifications and safeguards on the following schedule:
Timing | What happens |
|---|---|
2.5 years before CA expiration | You receive an expiry reminder notification |
2 years before CA expiration | A successor CA is automatically appended, if you haven't created one yourself |
Shortly after successor CA distribution completes | You receive a notification that distribution is complete |
60 days before auto-activation | You receive a warning that automatic activation is coming soon |
6 months before CA expiration | The successor CA is automatically activated, if you haven't activated it yourself |
45 days before CA expiration | A final automatic activation occurs, even if you had previously rolled back |
Note that clusters created in 2018–2019 follow an adjusted notification schedule, because applying the standard schedule as-is would put the calculated start date before this feature even became available. Rather than waiting for a notification to arrive, it's worth checking your cluster's CA expiration with
describe-certificate-authority now.Let AWS Handle It, or Take the Lead?
There are broadly two ways to approach CA rotation.
The first is to let AWS's automation handle it. If you take no action, a successor CA is automatically appended two years before expiration, and automatically activated six months before expiration. For components that AWS fully manages — EKS Auto Mode instances and Fargate nodes — this automated path is all you need.
The second is to take the lead yourself. Using
create-certificate-authority , you can append a successor CA at any time, without waiting for AWS's automated timeline. Starting earlier gives you more breathing room to identify and update your worker nodes and external clients.It's worth keeping in mind what these two approaches actually affect: not whether the cluster stays up, but whether individual clients keep their connectivity. AWS's automation is a safeguard for the availability of the cluster (the control plane) itself, and that holds true whether you take the automated path or the self-directed one. Updating the worker nodes and external clients you manage, on the other hand, is always your responsibility, regardless of which path you choose. If those updates aren't done by the time the successor CA is activated, the cluster itself keeps running, but any client that hasn't been updated loses its connection to the API server.
In environments where multiple teams or CI/CD pipelines connect to the same cluster, simply identifying all the external clients can take real time. Starting early, on your own terms, tends to be a lot less stressful than racing AWS's automated timeline.
Getting Started
This section walks through performing CA rotation with the AWS CLI. All you need is AWS CLI version 2.x or later and the same IAM permissions you already use to call the EKS API — no additional permissions are required specifically for CA rotation.
Check Your Active CA
Start by checking the CA currently configured on your cluster.
aws eks list-certificate-authorities --cluster-name <cluster-name> --region <region>
Then check that CA's expiration date.
aws eks describe-certificate-authority \ --cluster-name <cluster-name> \ --certificate-authority-id <ca-id> \ --region <region>
The expiration date appears in the response under
validity.notAfter .Append a Successor CA
When you're ready, append a successor CA.
aws eks create-certificate-authority --cluster-name <cluster-name> --region <region>
The response includes an
updateId you can use to track progress.aws eks describe-update --name <cluster-name> --update-id <update-id> --region <region>
Wait until the status shows
Successful .Verify Distribution Status
Running
list-certificate-authorities again now shows two CAs. You can't proceed to activation until the successor CA's distributionStatus moves from IN_PROGRESS to COMPLETE .aws eks list-certificate-authorities --cluster-name <cluster-name> --region <region>
Update Your kubeconfig
Once distribution is complete, update your own kubeconfig.
aws eks update-kubeconfig --name <cluster-name> --region <region>
This ensures your kubectl commands keep working after the successor CA is activated.
Update Worker Nodes and External Clients
This is where "your responsibility" comes into play. What you need to do depends on the type of worker node.
Worker node type | Action required |
|---|---|
Managed node groups | Run aws eks update-nodegroup-version to trigger a rolling replacement of nodes |
Karpenter-controlled nodes | Nodes are replaced automatically if drift detection is enabled; otherwise, use kubectl cordon / kubectl drain to trigger replacement manually |
Self-managed nodes | Update the CA data in your launch template, then perform a rolling replacement through your Auto Scaling group |
AWS Fargate | No action required — pods are recycled naturally through AWS's patching process |
EKS Auto Mode | No action required — AWS updates these automatically |
For external clients (kubeconfig files, CI/CD pipelines, monitoring tools, and so on), update their trust configuration with the latest CA data, retrieved as follows:
aws eks describe-cluster \ --name <cluster-name> --region <region> \ --query 'cluster.certificateAuthority.data' --output text
Activate the Successor CA
Once your worker nodes and external clients have been updated, activate the successor CA.
aws eks activate-certificate-authority \ --cluster-name <cluster-name> \ --certificate-authority-id <successor-ca-id> \ --region <region>
After activation, the cluster issues certificates signed by the successor CA. Running
kubectl get nodes afterward is a good way to confirm connectivity is intact.Using the Console
You can perform the same operations from the Amazon EKS console. It shows your CA's status and expiration, along with a guided view of rotation progress — append, distribution, worker node and external client updates, activation, and outgoing CA retirement — which can be easier to follow if you're less comfortable with the CLI.
If You Activate Before Everything Is Updated
If the successor CA ends up activated before you've finished updating your worker nodes and external clients, you can roll back to the previous CA.
aws eks activate-certificate-authority \ --cluster-name <cluster-name> \ --certificate-authority-id <previous-ca-id> \ --region <region>
You can check whether rollback is available via the
rollbackAvailable field returned by describe-certificate-authority . Keep in mind that rollback is only available before the final automatic activation, which occurs 45 days before CA expiration.
Summary
Amazon EKS CA rotation addresses a problem that customers previously had to manage entirely on their own: CA expiration. It provides a managed lifecycle and automated safeguards, and its three-stage process — appending, distributing, and activating a successor CA, connected by a dual trust period — lets you migrate to a new CA incrementally, without taking the cluster down.
That said, what AWS's automation guarantees is the availability of the cluster itself. Updating your worker nodes and external clients remains work you need to do. A good place to start is checking your cluster's CA expiration with
describe-certificate-authority , then beginning the work of identifying your external clients.
References
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article