EKS - Vertical Pod Autoscaling (VPA)

Step-01: Introduction

• The Kubernetes Vertical Pod Autoscaler automatically adjusts the CPU and memory reservations for your pods to help "right size" your applications.
• This adjustment can improve cluster resource utilization and free up CPU and memory for other pods.

Step-02: Pre-requisite - Metric server

• Metric server should be installed & ready, we have done that as part of HPA

Step-03: Deploy the Vertical Pod Autoscaler (VPA)

# Clone Repo
git clone https://github.com/kubernetes/autoscaler.git

# Navigate to VPA
cd autoscaler/vertical-pod-autoscaler/

# Uninstall VPA (if we are using old one)
./hack/vpa-down.sh

# Install new version of VPA
./hack/vpa-up.sh

# Verify VPA Pods
kubectl get pods -n kube-system

Step-04: Review & Deploy our Application Manifests (Deployment & Service)

• Make a note of resources we have defined spec.containers.resources.requests.
• limits we define in VPA definition

        resources:
          requests:
            cpu: "5m"       
            memory: "5Mi"   

• Deploy

# Deploy Application
kubectl apply -f kube-manifests/01-VPA-DemoApplication.yml

# List Pods, Deploy & Service
kubectl get pod,svc,deploy

# Describe Pod
kubectl describe pod <pod-name>

# Access Application (If our NodeGroup is in Public Subnet, if not ignore)
kubectl get nodes -o wide
http://<Worker-Node-Public-IP>:31232

Step-05: Create & Deploy VPA manifest

Create VPA Manifest

• Create a VPA manifest for our above Application which we deployed just now.

apiVersion: "autoscaling.k8s.io/v1beta2"
kind: VerticalPodAutoscaler
metadata:
  name: kubengix-vpa
spec:
  targetRef:
    apiVersion: "apps/v1"
    kind: Deployment
    name: vpa-demo-deployment
  resourcePolicy:
    containerPolicies:
      - containerName: '*'
        minAllowed:
          cpu: 5m
          memory: 5Mi
        maxAllowed:
          cpu: 1
          memory: 500Mi
        controlledResources: ["cpu", "memory"]

Deploy VPA Manifest

# Deploy
kubectl apply -f kube-manifests/02-VPA-Manifest.yml

# List VPA
kubectl get vpa

# Describe VPA
kubectl describe vpa kubengix-vpa

Step-06: Generate Load

• Open 3 more new terminals and execute below 3 load generation commands

# Terminal 1 - List and watch pods
kubectl get pods -w

# Terminal 2 - Generate Load
kubectl run --generator=run-pod/v1 apache-bench -i --tty --rm --image=httpd -- ab -n 500000 -c 1000 http://vpa-demo-service-nginx.default.svc.cluster.local/

# Terminal 3 - Generate Load
kubectl run --generator=run-pod/v1 apache-bench2 -i --tty --rm --image=httpd -- ab -n 500000 -c 1000 http://vpa-demo-service-nginx.default.svc.cluster.local/

# Terminal 4 - Generate Load
kubectl run --generator=run-pod/v1 apache-bench3 -i --tty --rm --image=httpd -- ab -n 500000 -c 1000 http://vpa-demo-service-nginx.default.svc.cluster.local/

Step-07: Describe pods which were re-launched by VPA Updater

# List Pods
kubectl get pods

# Describe pods
kubectl describe pod <recently-relaunched-pod>

Step-08: Important Nodes about VPA:

1. VPA Updater can re-launch new pod with updated CPU and Memory when you atleast have 2 pods in a deployment.
2. If we have only one pod, unless we manually delete that pod, it will not launch new pod with VPA recommended CPU and memory considerign the application availability scenario.

Step-09: Clean-Up

kubectl delete -f kube-manifests/