flavono123 / eks-better-practice

AWS EKS Better Practice with Terraform

Update frequently(break points)

Kubeconfig(Pre-requisite)

• awscli
• kubectx

export AWS_PROFLIE=YOUR_PROFILE
export CLUSTER_NAME=YOUR_CLUSTER_NAME

# Cleanup
# caveat: the cluster and context would be removed but the USER would NOT
k ctx -d $AWS_PROFILE@$CLUSTER_NAME
k config delete-cluster $CLUSTER_NAME

# TODO: more elegant way e.g. cli, krew?
# Set a simple context and user name,
# each with CLUSTER_NAME and AWS_PROFLIE, not the verbose arn
aws eks update-kubeconfig --name $CLUSTER_NAME --alias $AWS_PROFILE@$CLUSTER_NAME --user-alias $AWS_PROFILE
# Set a simple cluster name over the verbose arn
sed -i s%$(aws eks describe-cluster --name $CLUSTER_NAME | jq -r .cluster.arn | awk -F/ '{ print $1"/" }')%% -- $HOME/.kube/config

Roadmap

Implement EKS workshop's each part

• Fundamentals
  • Fargate
  • AWS Load Balancer Controller
  • Storage
    • EBS
    • EFS
• Autoscaling
  • Karpenter
  • Horizontal Pod Autoscaler
  • Cluster Proportional Autoscaler

Diagrams

Two-steps of Provisioning

• One provisionig with Terraform
  • VPC
  • EKS cluster
  • Karpenter(AWS' things)
  • Helm
    • Karpenter
    • ArgoCD
• Two provisioning with ArgoCD
  • Karpenter things
    • AWSNodeTemplates
    • Provisioners
  • Sample applications

How to connect two steps?

• Terraform outputs should be passed Helm(ArgoCD) values.