EKS

Amazon Elastic Kubernetes Service (EKS) runs the Kubernetes control plane for you, so you do not operate the API server or etcd. You remain responsible for the compute that runs your workloads, the networking around it, and the applications themselves.

How it works

You create a cluster and AWS operates its control plane. Compute comes in three forms: managed node groups, self-managed nodes on EC2, and Fargate profiles. Managed node groups automate provisioning and lifecycle of EC2 nodes, but you still decide when to update node images and Kubernetes versions. Nodes can also be scaled dynamically with Karpenter.

Common integrations include the AWS Load Balancer Controller for ALB and NLB, the EBS CSI driver for persistent volumes, and IAM roles for service accounts (IRSA), which give pods scoped AWS permissions instead of sharing the node role.

A practical example

A team creates a cluster with Terraform, adds a managed node group across private subnets, installs the AWS Load Balancer Controller, and annotates a Kubernetes service account with an IAM role so a pod can read one S3 bucket. Access is granted with RBAC inside the cluster and IAM outside it.

Cost

AWS lists the control plane at $0.10 per cluster per hour under standard Kubernetes version support, and $0.60 per hour under extended support, at the time of writing. Worker nodes, storage, load balancers and data transfer are billed separately and are usually the larger share. Check the current EKS pricing page before budgeting.

Common mistakes

• Letting the cluster version fall behind. Older versions move into extended support, which costs more, so plan regular upgrades.
• Sharing the node IAM role with every pod instead of using IRSA.
• Running out of pod IP addresses in small subnets.
• Omitting resource requests and limits, which makes autoscaling and bin-packing unreliable.
• Exposing the API endpoint more widely than needed.

When to use it and when not to

EKS fits teams that want standard Kubernetes on AWS with deep integration into IAM, VPC and load balancing. It is a poor fit for a handful of simple containers, where a simpler service may cost less to run and maintain. If you run across clouds, compare it with AKS and GKE. Our AWS EKS consulting page covers design, upgrades and cost work.

See also

• Kubernetes
• AKS
• Fargate
• Karpenter
• EKS cluster setup