https://drive.google.com/file/d/15MkFUZqIGv7FbcPxh02Na9ryhe2dBAcE/view?usp=sharing

🔍 HPA v1: Simple CPU Autoscaling

# hpa-for-autoscaler-deployment.yaml
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: k8s-autoscaler
spec:
  minReplicas: 2
  maxReplicas: 10
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: k8s-autoscaler
  targetCPUUtilizationPercentage: 10  # ← Scale when CPU > 10% of request

🔑 Key Points:

• CPU-only
• targetCPUUtilizationPercentage = % of requested CPU
• Simple, but limited

⚠️ Why 10%?

Very aggressive scaling!

• If Pod requests 200m CPU → scales when usage > 20m
• Good for bursty workloads, but may cause flapping

🔍 HPA v2: Multi-Metric & Flexible

# hpa-for-deployment-v2.yaml
apiVersion: autoscaling/v2beta2  # ← Note: v2beta2 (or v2 in 1.23+)
kind: HorizontalPodAutoscaler
metadata:
  name: my-app
spec:
  minReplicas: 1
  maxReplicas: 5
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 66  # ← Scale when CPU > 66% of request

🔑 Key Improvements:

• metrics array → supports CPU, memory, custom, external
• averageUtilization → clearer than v1’s percentage
• Multiple metrics → scale on CPU OR memory OR both

💡 HPA v2 Structure:

metrics:
- type: Resource       # CPU/memory
  resource: ...
- type: Pods           # Custom per-pod metric
  pods: ...
- type: Object         # Metric from another object
  object: ...
- type: External       # External metric (e.g., queue depth)
  external: ...

📌 Critical Notes for k3s

1. API Version:

   • Kubernetes 1.23+ → use autoscaling/v2 (stable)
   • Older → autoscaling/v2beta2 → k3s v1.28+ supports v2

2. Metrics Server:

   Must be installed (as we verified earlier).

3. Resource Requests:

   Still required for CPU/memory metrics.

🧪 k3s Lab: Compare HPA v1 vs v2

🔧 Step 1: Deploy Target Application

💡 Use your existing deployment-for-autoscaler.yaml (with run: k8s-autoscaler)

kubectl apply -f deployment-for-autoscaler.yaml