Ping Server for Kubernetes

A simple Express.js "ping" server designed to be deployed and scaled on Kubernetes using kind.

Prerequisites

• Docker Desktop (running)
• kind (brew install kind on macOS)
• kubectl (brew install kubectl on macOS)
• Node.js and npm

Quick Start

1. Setup the environment:

   chmod +x setup.sh deploy.sh fix-metrics-server.sh test-autoscaling.sh
   ./setup.sh

2. Deploy to kind cluster:

   ./deploy.sh

3. If metrics server has issues (common with kind):

   ./fix-metrics-server.sh

4. Test the service:

   curl http://localhost:30080/ping

Project Structure

.
├── server.js              # Express.js server
├── package.json           # Node.js dependencies
├── Dockerfile             # Container image definition
├── kind-config.yaml       # Kind cluster configuration
├── setup.sh              # Environment setup script
├── deploy.sh              # Deployment script
├── fix-metrics-server.sh  # Metrics server troubleshooting
├── test-autoscaling.sh    # Autoscaling load test
└── k8s/
    ├── deployment.yaml    # Kubernetes deployment
    ├── service.yaml       # Kubernetes services
    └── hpa.yaml          # Horizontal Pod Autoscaler

API Endpoints

• GET / - Server information
• GET /ping - Returns "pong" with server details
• GET /health - Health check endpoint
• GET /stress?duration=5000 - CPU stress test for autoscaling (duration in ms, max 30s)

Autoscaling Features

Horizontal Pod Autoscaler (HPA)

• Min Replicas: 1
• Max Replicas: 15
• CPU Threshold: 60% average utilization
• Memory Threshold: 70% average utilization
• Scale Up: Fast (15s stabilization, up to 100% increase)
• Scale Down: Conservative (60s stabilization, max 50% decrease)

Testing Autoscaling

# Make scripts executable
chmod +x test-autoscaling.sh

# Run load test to trigger autoscaling
./test-autoscaling.sh

The load test will:

1. Generate high-frequency ping requests
2. Call CPU stress endpoints
3. Monitor scaling in real-time
4. Save scaling metrics to scaling_log.csv

Scaling

The server automatically scales between 1-15 replicas based on:

• CPU usage > 60%
• Memory usage > 70%

Manual scaling:

kubectl scale deployment ping-server --replicas=5

Real-time Autoscaling Monitoring

# Watch HPA status
kubectl get hpa ping-server-hpa --watch

# Monitor pod scaling
watch kubectl get pods -l app=ping-server

Monitoring

# Check pods
kubectl get pods -l app=ping-server

# Check HPA status
kubectl get hpa ping-server-hpa

# View logs
kubectl logs -l app=ping-server --tail=50

# Port forward for local access
kubectl port-forward svc/ping-server-service 8080:80

Cleanup

# Delete deployment only
kubectl delete -f k8s/

# Delete entire kind cluster
kind delete cluster --name ping-server-cluster

Development

Run locally for development:

npm install
npm start
# Server runs on http://localhost:3000