This module deploys a complete publicly accessible EKS cluster in an empty AWS account. All that is required is to have an AWS account and user created, awscli configured, and kubectl installed.

By leveraging existing EKS and VPC modules from the terraform modules registry, this module itself is kept relatively simple.

Use the following terraform.auto.tfvars to deploy a VPC, EIP, NAT gateway, routes, public/private subnets, security groups, NAT gateway, IAM roles & policies, EKS backplane, and two worker groups (one using on-demand instances, the other on spot instances).

NOTE: If you wish to deploy to an existing VPC, the VPC module can be disabled by adding create_vpc = false in the .tfvars and then existing VPC referenced using the existing_vpc_* variables. Note that all resources that would have been created by the VPC module must be present.

# terraform.auto.tfvars

region = "us-east-1"

cluster_name       = "demo"
kubernetes_version = "1.24"
ami_type           = "BOTTLEROCKET_x86_64"

vpc_cidr        = "10.0.0.0/16"
private_subnets = ["10.0.1.0/24", "10.0.2.0/24", "10.0.3.0/24"]
public_subnets  = ["10.0.101.0/24", "10.0.102.0/24", "10.0.103.0/24"]

eks_managed_node_groups = [
  {
    name           = "demo-on-demand"
    instance_types = ["t3.small"]

    min_size     = 1
    max_size     = 1
    desired_size = 1
  },
  {
    name           = "demo-spot"
    instance_types = ["t3.small"]

    capacity_type = "SPOT"

    min_size     = 1
    max_size     = 1
    desired_size = 1
  },
]

Once the file has been added, run terraform plan to see what will be deployed. Then run terraform apply -auto-approve to deploy the infrastructure. This will take around 10-15 minutes.

After the apply is complete, run the following to configure the Kubernetes client:

aws eks update-kubeconfig --name $(terraform output -raw cluster_name)

The nodes and pods can then be checked by running the following:

kubectl get nodes
kubectl get pod -A

Finally, run terraform destroy -auto-approve when you are ready to destroy all the created resources.

list(object({
    name = string

    instance_types = list(string)

    min_size     = number
    max_size     = number
    desired_size = number

    capacity_type = optional(string, "ON_DEMAND") # Alternatively "SPOT"
    labels        = optional(map(string), {})

  }))