EKS Architecture Overview
What you'll learn: How Amazon EKS separates responsibilities between AWS-managed control plane components and your own worker infrastructure.
Amazon EKS (Elastic Kubernetes Service) implements a shared responsibility model for Kubernetes infrastructure. Unlike self-managed clusters where you operate everything, EKS splits the architecture into two distinct zones.
The AWS-Managed Control Plane
AWS runs the entire Kubernetes control plane for you—the API server, etcd datastore, scheduler, and controller managers. These components live in AWS-owned infrastructure, not your account. AWS handles:
- High availability (spanning multiple AZs automatically)
- Security patches and version upgrades
- Scaling the control plane based on cluster load
- Backups and disaster recovery
You interact with this control plane through the standard Kubernetes API, but you never see or manage the underlying machines.
Worker Nodes in Your VPC
Your worker nodes (EC2 instances or Fargate pods) run in your own VPC. These are the machines that actually execute your containerized workloads. You maintain control over:
- Instance types and sizing
- Operating system patches (for EC2 workers)
- VPC networking and security groups
- IAM roles for workload permissions
The Connection
Worker nodes authenticate to the AWS-managed control plane using IAM and specialized bootstrap logic. The kubelet on each node registers with the API server across the AWS network boundary, but workload traffic stays entirely within your VPC.
Think of it like renting a high-end apartment building: AWS maintains the elevators, security systems, and infrastructure (control plane), while you furnish and manage your individual units (worker nodes).
Key Takeaway: EKS separates the control plane (AWS-managed, invisible to you) from worker nodes (your VPC, your responsibility), giving you managed Kubernetes without losing infrastructure control.