AutoScaler

A Go web application demonstrating Kubernetes Horizontal Pod Autoscaling (HPA) with Prometheus metrics.

Endpoints

Method	Path	Description
GET	/health	Health check
GET	/metrics	Prometheus metrics
POST	/load	Generate CPU load to trigger HPA scaling

POST /load

{ "amount": 4, "duration": 30 }

• amount — number of CPU-bound goroutines (1–64, default: number of CPUs)
• duration — how long to run in seconds (1–300, default: 10)

Prerequisites

• Docker
• minikube
• kubectl
• Helm
• k6

Quick Start

1. Start minikube

minikube start
minikube addons enable metrics-server
minikube addons enable ingress

2. Build and load the image

make load-image

3. Deploy the application

make deploy

Verify the app is running via port-forward:

# Terminal 1 — keep open
kubectl port-forward svc/auto-scaler-service 8080:80

# Terminal 2
curl http://localhost:8080/health

Note (WSL2 / Docker driver): Ingress is not directly reachable on WSL2 with the Docker driver because NodePorts are not exposed from the minikube container. Use kubectl port-forward as shown above, or run minikube tunnel in a separate terminal to enable Ingress access via http://auto-scaler.local.

4. Install Prometheus + Grafana

make setup-monitoring

Wait until all pods are running:

kubectl get pods -w

Access Grafana (credentials: admin / prom-operator):

minikube service kube-prometheus-grafana

To get password for admin account in Grafana:

kubectl --namespace default get secrets kube-prometheus-grafana -o jsonpath="{.data.admin-password}" | base64 -d ; echo

Note (WSL2): If the browser doesn't open automatically, run minikube service kube-prometheus-grafana --url to get the URL and open it manually.

5. Run load test

Open three terminals:

# Terminal 1 — port-forward (keep open)
kubectl port-forward svc/auto-scaler-service 8080:80

# Terminal 2 — watch HPA react in real time
make watch-hpa

# Terminal 3 — run the load test
BASE_URL=http://localhost:8080 make load-test

Expected behavior: CPU usage rises above 50% → HPA scales pods up to max 5 → load ends → pods scale back down after ~60s.

You can also see the CPU usage and amount of replicas using following command

kubectl get hpa auto-scaler-hpa

Inside Grafana here are the main dashboards:

• Kubernetes / Compute Resources / Workload
• Kubernetes / Compute Resources / Namespace (Workloads)

There you can track number of pods, CPU usage and other

Metrics

The app exposes the following Prometheus metrics:

Metric	Type	Description
http_requests_total	Counter	Total requests by method, path, status
http_request_duration_seconds	Histogram	Request latency
http_requests_dropped_total	Counter	Requests canceled or timed out

Useful PromQL queries in Grafana (Explore → Prometheus):

# Replica count over time
kube_deployment_status_replicas{deployment="auto-scaler"}

# Requests per second
rate(http_requests_total[1m])

# Average response time
rate(http_request_duration_seconds_sum[1m]) / rate(http_request_duration_seconds_count[1m])

Project Structure

.
├── cmd/
│   └── main.go
├── internal/
│   ├── handlers/
│   │   ├── dto.go
│   │   ├── health.go
│   │   ├── load.go
│   │   └── metrics.go
│   └── util/
│       └── json.go
├── deploy/k8s/
│   ├── deployment.yaml
│   ├── service.yaml
│   ├── ingress.yaml
│   ├── hpa.yaml
│   └── servicemonitor.yaml
├── load-test/
│   └── load-test.js
├── Dockerfile
└── Makefile