Introduction

This repo deploys a highly available EKS cluster with an Autoscaling Group in region us-east-1 across 3 different Availability Zones.

Since I've used AWS Kodekloud Playgrounds to elaborate the contents of this repo, some actions and resources were restricted or had technical limitations.

Actions such as using specific terraform modules, selecting instance types other than t2.micro, setting custom EKS cluster naming or using an up to date EKS optimized ami, were either restricted or not possible.

📁 /
 ├──📁 terraform/
 |    ├── aws-auth-cm.yaml
 |    ├── controlplane.tf
 |    ├── dataplane.tf
 |    ├── iam.tf
 |    ├── networking.tf
 |    ├── providers.tf
 |    ├── variables.tf
 |    ├── userdata.sh
 ├── script.sh
 ├── kubelet.service
 ├── README.md

Technologies:

• 
  • 
  • 
  • 
  • 
• 
• 
• 
• 

Diagram

Prerequisites

• AWS CLI with your AWS account.
• Terraform with an access key.
• kubectl

Run $ ./script.sh to verify if you meet all the prerequisites. The script must be executable $ sudo chmod u+x script.sh.

The script checks if:

• AWS CLI is installed and downdloads if it's not.
• Terraform is installed and downdloads if it's not.
• kubectl is installed and downdloads if it's not.
• Creates an AWS S3 Bucket with versioning enabled to store the Terraform State. Feel free to modify the var BUCKET but remember it must match the name of the bucket in the S3 backend, within the providers.tf file
• Generates a key pair dataplane-kp.pem and dataplane-kp.pem.pub for the EC2 Autoscaling Group.

Deployment

Make sure that the script generates a key pair. Alternatively you can generate a new key pair with the commands below.

• $ ssh-keygen -t rsa -N "" -f ${HOME}/dataplane-kp.pem
• $ export TF_VAR_dataplane_public_key=$(cat "${HOME}/dataplane-kp.pem.pub")

Navigate into the /terraform directory and start the terraform deployment cycle. It will take time, so grab a cup of coffee and let it work for 8 minutes or so.

• $ terraform init
• $ terraform plan
• $ terraform apply

Once the deployment is completed, you need to copy the output value dataplane-role-arn shown in the console and paste it in the aws-auth-cm.yaml > rolearn field.

Update kube-config to access the EKS control plane with:

• $ aws eks update-kubeconfig --region us-east-1 --name eks-demo

Join the dataplane nodes with the Controlplane with kubectl and allow EKS 60 seconds to detect the autoscaling group.

• $ kubectl apply -f aws-auth-cm.yaml

Observations

• The worker nodes of the dataplane use an EKS optimized ami that comes with bootstrap.sh pre installed. However, since this EKS deployment is using kubernetes 1.32 the ami needs an upgrade.

• The script userdata.sh that is passed to the EC2 launch template contains a workaround to make communications between the dataplane and EKS controlplane possible.

• The AWS LoadBalancer Controller creates an AWS Application Load Balancer (ALB) when you create a Kubernetes Ingress resource and creates Network Load Balancer (NLB) when you create a Kubernetes service of type LoadBalancer.

Future plans

In time, I will refactor the contents of this repo and implement them as part of a CI/CD pipeline using Jenkins and hopefully switch everything to Terraform modules.

Terraform Documentation

Networking	IAM	Compute
Resource: aws_vpc	Resource: aws_iam_role	Resource: aws_eks_cluster
Resource: aws_subnet	Resource: aws_iam_role_policy	Resource: aws_launch_template
Resource: aws_default_vpc	Resource: aws_iam_user_policy_attachment	Resource: aws_autoscaling_group
Resource: aws_default_subnet	--	Resource: aws_key_pair
Resource: aws_security_group	--	--
Resource: aws_security_group_rule	--	--
Resource: aws_default_security_group	--	--
Resource: aws_internet_gateway	--	--
Resource: aws_route_table	--	--
Resource: aws_route_table_association	--	--
Resource: aws_default_route_table	--	--
Resource: aws_vpc_peering_connection	--	--

AWS Documentation

Networking	Compute	Misc.
IPv4 VPC CIDR blocks	Prerequisites for EC2 Instance Connect	amazon-elastic-kubernetes-service-course
EKS Loadbalancing	Customize managed nodes with launch templates	Create EKS Cluster with Terraform
Route internet traffic with AWS Load Balancer Controller	How do I get my worker nodes to join my Amazon EKS cluster?	--
--	How do I troubleshoot Amazon EKS managed node group creation failures?	--
--	How do I troubleshoot Amazon EKS managed node group creation failures?	--