Skip to content

Lesson 01 — Amazon EKS Architecture

By the end of this lesson, you will be able to:

  • Explain the architecture of Amazon EKS
  • Understand the components of a Kubernetes cluster running on AWS
  • Differentiate between the Kubernetes Control Plane and Worker Nodes
  • Understand how Amazon EKS manages Kubernetes
  • Explain the networking architecture of Amazon EKS
  • Understand EKS node groups and compute options
  • Design secure enterprise EKS architectures
  • Understand multi-account and multi-region EKS deployments
  • Identify AWS services commonly integrated with Amazon EKS
  • Apply Amazon EKS architecture best practices

Amazon Elastic Kubernetes Service (Amazon EKS) is one of the most widely adopted managed Kubernetes platforms in the cloud.

Many organisations use Amazon EKS to run:

  • Enterprise web applications
  • Banking platforms
  • Payment systems
  • AI and Machine Learning workloads
  • Containerised microservices
  • DevSecOps pipelines
  • Security platforms
  • Large-scale SaaS applications

As a Cloud Security Engineer, understanding Amazon EKS architecture is essential because every security control depends on understanding how the platform is built.

You cannot secure:

  • Kubernetes API
  • Worker Nodes
  • Networking
  • IAM
  • Storage
  • Secrets
  • Logging
  • Monitoring

without first understanding how they fit together.

Amazon Elastic Kubernetes Service (Amazon EKS) is a fully managed Kubernetes service provided by AWS.

AWS manages the Kubernetes control plane while customers manage the worker nodes, workloads and Kubernetes resources.

AWS
Managed Kubernetes Control Plane
+
Customer
Applications
Worker Nodes
Networking
Security
Policies

Amazon EKS removes much of the operational burden associated with managing Kubernetes clusters.

Users
AWS IAM
Amazon EKS API Endpoint
Managed Kubernetes Control Plane
├── API Server
├── Scheduler
├── Controller Manager
└── etcd
Worker Nodes
Pods
Applications

This separation between the managed control plane and customer-managed worker nodes is one of the defining characteristics of Amazon EKS.

Amazon EKS follows the AWS Shared Responsibility Model.

AWS manages:

  • Kubernetes control plane
  • etcd infrastructure
  • API Server availability
  • Scheduler
  • Controller Manager
  • Managed service infrastructure
  • Control plane scaling
  • Control plane patching
  • High availability of the managed service

Customers manage:

  • Worker Nodes
  • Managed Node Groups configuration
  • Self-managed nodes
  • Kubernetes workloads
  • RBAC
  • IAM
  • Networking
  • Security Groups
  • Secrets
  • Applications
  • Container Images
  • Logging
  • Monitoring
  • Compliance
AWS
└── Control Plane
Customer
├── Nodes
├── Pods
├── Applications
├── Security
└── Governance

An Amazon EKS cluster consists of several major components.

Amazon EKS Cluster
├── Managed Control Plane
├── Worker Nodes
├── Kubernetes Pods
├── Services
├── Networking
├── Storage
├── IAM
├── Logging
└── Monitoring

Each component plays an important role.

The Kubernetes Control Plane is the brain of the cluster.

AWS manages this entire layer.

It includes:

  • Kubernetes API Server
  • Scheduler
  • Controller Manager
  • etcd
  • Cloud Controller integration

Applications never interact directly with etcd.

Instead:

kubectl
API Server
etcd

The API Server validates and stores all Kubernetes objects.

The API Server is the central communication point.

Everything interacts through it.

Examples include:

  • kubectl
  • CI/CD pipelines
  • Controllers
  • Operators
  • Admission Controllers
  • Kubernetes Dashboard
  • Automation tools
Developer
kubectl
API Server
Cluster

Every request passes through:

  • Authentication
  • Authorisation
  • Admission Controllers
  • Validation

before being stored.

The Scheduler determines where Pods should run.

It considers:

  • CPU availability
  • Memory availability
  • Resource requests
  • Node affinity
  • Taints
  • Tolerations
  • Topology constraints
  • Scheduling policies
New Pod
Scheduler
Best Worker Node

The Controller Manager maintains the desired cluster state.

Controllers include:

  • Deployment Controller
  • ReplicaSet Controller
  • Node Controller
  • Job Controller
  • Endpoint Controller

Example:

Desired replicas:

3 Pods

Current state:

2 Pods

Controller Manager automatically creates:

1 New Pod

etcd is Kubernetes’ distributed key-value database.

It stores:

  • Pods
  • Deployments
  • Secrets
  • ConfigMaps
  • Services
  • Namespaces
  • RBAC
  • Policies

AWS manages etcd within Amazon EKS.

Customers never access etcd directly.

Worker Nodes run applications.

They provide:

  • CPU
  • Memory
  • Storage
  • Network connectivity

Worker Nodes contain:

Worker Node
├── kubelet
├── Container Runtime
├── kube-proxy
└── Pods

kubelet communicates with the Kubernetes API Server.

Responsibilities include:

  • Starting containers
  • Monitoring Pods
  • Reporting node status
  • Running health checks
  • Managing volumes
API Server
kubelet
Containers

The Container Runtime executes containers.

Modern Amazon EKS clusters use:

  • containerd

The runtime:

  • Pulls images
  • Starts containers
  • Stops containers
  • Reports container status

kube-proxy manages Kubernetes networking.

Responsibilities include:

  • Service routing
  • ClusterIP
  • NodePort
  • Load balancing
  • iptables rules
  • IPVS rules (where applicable)

It enables Pods to communicate through Kubernetes Services.

Pods are the smallest deployable Kubernetes object.

A Pod contains:

  • One or more containers
  • Shared networking
  • Shared storage
  • Shared lifecycle

Example:

Pod
├── Application Container
└── Logging Sidecar

Networking is built on Amazon VPC.

Amazon VPC
Subnets
Worker Nodes
Pods

Every Pod receives its own IP address.

Networking components include:

  • Amazon VPC
  • Security Groups
  • Route Tables
  • Elastic Network Interfaces (ENIs)
  • Kubernetes Services
  • Network Policies

Amazon EKS uses the Amazon VPC CNI plugin.

Benefits include:

  • Native VPC networking
  • Pod IP addresses from VPC
  • Security Group integration
  • High performance
VPC
ENI
Pod IP

Each Worker Node receives one or more ENIs.

Each ENI provides IP addresses for Pods.

EC2 Instance
ENI
Multiple Pod IPs

The number of Pods depends on:

  • EC2 instance type
  • Number of ENIs
  • Available IP addresses

Services expose applications.

Common types include:

Service Type Purpose
ClusterIP Internal communication
NodePort Exposes through worker nodes
LoadBalancer AWS Load Balancer
ExternalName External DNS

Ingress manages HTTP and HTTPS routing.

Common implementations include:

  • AWS Load Balancer Controller
  • NGINX Ingress Controller
Internet
Application Load Balancer
Ingress
Service
Pods

Storage options include:

  • Amazon EBS
  • Amazon EFS
  • FSx
  • CSI Drivers
Pod
Persistent Volume
Amazon EBS

Storage is dynamically provisioned using StorageClasses.

Identity combines AWS IAM and Kubernetes RBAC.

User
AWS IAM
Amazon EKS Authentication
Kubernetes RBAC
Cluster Access

Both layers work together.

Authentication uses AWS IAM.

Examples include:

  • IAM Users
  • IAM Roles
  • IAM Identity Center
  • Federated identities

Authentication determines:

Who are you?

RBAC determines:

What can you do?

Examples:

  • Create Pods
  • Delete Deployments
  • Read Secrets
  • Manage Nodes

Authentication and Authorisation are separate.

Amazon EKS supports multiple compute models.

Option Description
Managed Node Groups AWS-managed EC2 worker nodes
Self-managed Nodes Customer-managed EC2 instances
AWS Fargate Serverless Pods

AWS manages:

  • Scaling
  • Updates
  • Lifecycle integration

Customers manage:

  • Instance types
  • Scaling configuration
  • Security Groups
  • IAM roles

Managed Node Groups are recommended for most production workloads.

Customers manage:

  • EC2 lifecycle
  • AMIs
  • Updates
  • Scaling
  • Patching

Provides maximum flexibility but requires greater operational effort.

AWS Fargate eliminates worker node management.

Pod
AWS Fargate
Serverless Compute

Suitable for:

  • Small workloads
  • Event-driven applications
  • Operational simplicity

Production EKS clusters span multiple Availability Zones.

Region
├── AZ-A
├── AZ-B
└── AZ-C
Worker Nodes
Pods

This improves:

  • Availability
  • Fault tolerance
  • Disaster resilience

Example:

AZ-1
Worker Nodes
AZ-2
Worker Nodes
AZ-3
Worker Nodes

Applications should run across multiple AZs.

Large organisations commonly use multiple AWS accounts.

AWS Organizations
├── Security Account
├── Logging Account
├── Shared Services
├── Development
├── Testing
└── Production

Each account may contain one or more EKS clusters.

Critical workloads may span Regions.

Primary Region
Amazon EKS
Disaster Recovery Region
Amazon EKS

Benefits include:

  • Disaster Recovery
  • Business Continuity
  • Reduced regional risk

Amazon EKS security includes:

IAM
RBAC
Pod Security
Network Policies
Security Groups
Encryption
Monitoring

Multiple security layers provide defence in depth.

Logging sources include:

  • API Server
  • Audit Logs
  • Authenticator Logs
  • Scheduler Logs
  • Controller Manager Logs
  • Application Logs
  • Node Logs
Amazon EKS
CloudWatch
SIEM

Monitoring typically includes:

  • Amazon CloudWatch
  • Prometheus
  • Grafana
  • Amazon Managed Prometheus
  • Amazon Managed Grafana

Metrics include:

  • CPU
  • Memory
  • Network
  • API latency
  • Pod health

Common AWS Services Integrated with Amazon EKS

Section titled “Common AWS Services Integrated with Amazon EKS”
AWS Service Purpose
IAM Authentication
VPC Networking
EC2 Worker Nodes
ECR Container Registry
CloudWatch Monitoring
CloudTrail Auditing
KMS Encryption
Secrets Manager Secrets
GuardDuty Threat Detection
Security Hub Security Findings
Inspector Vulnerability Scanning
AWS Organizations
AWS Account
Amazon VPC
Amazon EKS
├── Managed Control Plane
├── Managed Node Groups
├── Ingress
├── Applications
├── Logging
├── Monitoring
├── Security Controls
└── GitOps

This architecture is common across enterprise AWS environments.

Risk

Cluster outage during AZ failure.

Recommendation

Deploy worker nodes across multiple AZs.


Risk

Increased attack surface.

Recommendation

Use private endpoints where appropriate and restrict public access.


Risk

Resource contention and security complexity.

Recommendation

Separate workloads using namespaces, node groups or multiple clusters.


Risk

Limited visibility during incidents.

Recommendation

Enable EKS control plane logging.


Risk

Configuration drift.

Recommendation

Use Managed Node Groups and Infrastructure as Code.

  • Define business requirements
  • Select AWS Region
  • Design VPC
  • Plan Availability Zones
  • Choose node model
  • Create Amazon EKS cluster
  • Configure networking
  • Deploy Managed Node Groups
  • Configure IAM
  • Configure RBAC
  • Enable logging
  • Configure encryption
  • Deploy admission policies
  • Configure Pod Security
  • Configure Network Policies
  • Monitor cluster
  • Patch worker nodes
  • Upgrade Kubernetes
  • Scan images
  • Review access
  • Continuous compliance
  • Policy management
  • Audit logging
  • Incident response
  • Security monitoring

As a Cloud Security Engineer:

  • Use Managed Node Groups unless specific requirements justify self-managed nodes.
  • Deploy worker nodes across multiple Availability Zones.
  • Protect the Kubernetes API endpoint with least-privilege access.
  • Use AWS IAM together with Kubernetes RBAC.
  • Enable Amazon EKS control plane logging.
  • Integrate Amazon EKS with CloudTrail, GuardDuty and Security Hub.
  • Store container images in Amazon ECR.
  • Encrypt sensitive data using AWS KMS.
  • Use Infrastructure as Code for cluster provisioning.
  • Continuously monitor cluster health, security posture and compliance.

A multinational financial institution operates more than 300 Amazon EKS clusters across multiple AWS accounts supporting payment systems, customer banking portals and fraud detection services.

To ensure a secure and scalable platform, the organisation adopts a standard Amazon EKS architecture that includes:

  1. AWS Organizations to separate production, development, security and shared services accounts.
  2. Multi-AZ Amazon EKS clusters with Managed Node Groups for high availability.
  3. Private API endpoints with restricted administrative access through AWS IAM Identity Center and MFA.
  4. Amazon VPC CNI for native networking and dedicated node groups for sensitive workloads.
  5. Amazon ECR as the approved container registry with mandatory image scanning and signing.
  6. Pod Security Admission and Kyverno to enforce enterprise security policies.
  7. CloudWatch, CloudTrail, GuardDuty and Security Hub integrated into a central SIEM for continuous monitoring.
  8. GitOps and Infrastructure as Code to provision and manage every cluster consistently across environments.

This standardised architecture enables the organisation to scale securely while maintaining strong governance, operational consistency and regulatory compliance.

  • Amazon EKS is a managed Kubernetes service that separates AWS-managed control plane responsibilities from customer-managed workloads.
  • The Kubernetes Control Plane consists of the API Server, Scheduler, Controller Manager and etcd.
  • Worker Nodes run applications and contain kubelet, containerd and kube-proxy.
  • Amazon VPC provides native networking for Pods using the Amazon VPC CNI.
  • AWS IAM authenticates users while Kubernetes RBAC authorises actions.
  • Managed Node Groups are the recommended compute option for most production environments.
  • Multi-AZ and multi-account architectures improve resilience and governance.
  • Amazon EKS integrates with many AWS security services such as CloudTrail, GuardDuty, Security Hub, KMS and Inspector.
  • Enterprise Amazon EKS deployments should follow standardised architectures, Infrastructure as Code and defence-in-depth security principles.

1. Which components make up the Amazon EKS managed control plane?

Section titled “1. Which components make up the Amazon EKS managed control plane?”

Answer: The Kubernetes API Server, Scheduler, Controller Manager and etcd.

2. What is the primary responsibility of Worker Nodes?

Section titled “2. What is the primary responsibility of Worker Nodes?”

Answer: Worker Nodes provide compute resources and run Kubernetes Pods and containers.

3. Which AWS networking component allows Pods to receive native VPC IP addresses?

Section titled “3. Which AWS networking component allows Pods to receive native VPC IP addresses?”

Answer: The Amazon VPC CNI plugin.

4. What is the difference between AWS IAM and Kubernetes RBAC?

Section titled “4. What is the difference between AWS IAM and Kubernetes RBAC?”

Answer: AWS IAM authenticates users and workloads to Amazon EKS, while Kubernetes RBAC determines what authenticated identities are authorised to do within the cluster.

Section titled “5. Why are Multi-AZ deployments recommended for Amazon EKS?”

Answer: They improve availability, fault tolerance and resilience by distributing worker nodes across multiple Availability Zones.

In the next lesson, we will explore Amazon EKS Cluster Design, including production-ready cluster topologies, node group strategies, networking layouts and enterprise deployment patterns.

➡️ Next Lesson: Lesson 02 — Amazon EKS Cluster Design