PROVISION MANAGED NODES AUTOSCALE EKS WITH PULUMI

Requirement

1. AWS CLI installed
2. Configure AWS CLI
3. Pulumi installed
4. Pulumi logged in
5. Helm instaled
6. NPM installed
7. Kubectl installed

How To

1. Clone repository

git clone https://github.com/ludesdeveloper/pulumi-managed-nodes-autoscale-eks.git 

2. Change directory

cd pulumi-managed-nodes-autoscale-eks

3. Install dependencies

npm install

4. Initialize Pulumi stack

pulumi stack init

5. Set Pulumi region

pulumi config set aws:region ap-southeast-1

6. Provision EKS with pulumi

pulumi up --yes

Give name of your stack, (dev) or other name you prefered

7. Copy kubeconfig to your kubeconfig directory (warning, this will replace your kubeconfig file)

pulumi stack output kubeConfig > ~/.kube/config

8. Make sure your cluster is ready

kubectl get nodes

9. You can run script below to apply cluster autoscaler or manually copy paste line inside init-autoscale-cluster.sh script

./init-autoscale-cluster.sh

Before Testing

1. Usually when testing, I open 3 tabs terminal to check, and every check consist of : auto scale log

kubectl -n kube-system logs -f deployment.apps/cluster-autoscaler

2. Check Nodes

watch kubectl get nodes

3. Check Pods

watch kubectl get pods -o wide

Testing

1. Apply nginx-deployment

kubectl apply -f nginx-deployment.yaml

2. Scale up nginx pod

kubectl scale deployment/nginx-deployment --replicas=20

You will see pending in your pod, and kubernetes will trigger scale up for nodes, wait for a while. You also can check in your EC2 dashboard AWS, AWS will try to create new instance

3. Scale down nginx pod

kubectl scale deployment/nginx-deployment --replicas=3

Scale down will takes time around 10 minutes, until cluster remove unnecessary nodes

Cleanup

1. Destroy pulumi

pulumi destroy --yes

2. Remove pulumi stack

pulumi stack rm dev

Type name of your stack

Source Articles

How to Scale Your Amazon EKS Cluster: EC2, Managed Node Groups, and Fargate

Amazon EKS Pulumi Crosswalk

Autoscaling - Amazon EKS