kubernetes-training

Horizontal Autoscaling

• Adding more instances

  The manual way

  kubectl scale deployment my-app --replica=3
  

  Using HPA

  Benefit

• Observe metrics
• Adds pods
• Balances thresholds
• TRack multiple metrics

  Imperative

• From the example above: when the cpu metrics goes above 50%, HPA modifies the number of pod replicas

  Declarative

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: php-apache
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: php-apache
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50
status:
  observedGeneration: 1
  lastScaleTime: <some-time>
  currentReplicas: 1
  desiredReplicas: 1
  currentMetrics:
  - type: Resource
    resource:
      name: cpu
      current:
        averageUtilization: 0
        averageValue: 0

• HPA comes built-in with Kubernetes since version 1.23

  Vertical Autoscaling

• Adding more resources to your existing applications

  The manual way

  kubectl edit deployment my-app
  

• Pod will be killed and recreated

  Using VPA

  Benefit

• Observe metrics
• Adjust pod resources
• Balances thresholds

Installation

1. VPA Recommender :
   • continuously monitor resource usage from the kubernetes metrics API
   • collects historical and live usage data for pods
   • provides optimal CPU and memory values
2. VPA Updater:
   • Detects pods running with sub-optimal resources and evicts them if an update is needed
   • The Updater needs information from the Recommender - VPA Admission Controller:
   • Interverne the pod creation process and uses the recommendations from the Recommender to apply the recommned resources

Declarative way

1. About updatePolicy:
   • Off: only Recommender working
   • Initial: only applied if there’s new pod created
   • Recreate: Updater would jump in and evict pods
   • Auto: similar to recreate for now, until in-place resize of pod is available

     Scaling Cluster Infra

     - Horizontal: adding more nodes to the cluster
     - Vertical scaling: adding resources to the existing nodes in the cluster ## Scaling Workloads
     -  horiz: getting more pods
     -  vertical: adding resources to the existing pods ![](../../../img/Pasted%20image%2020250621160254.png)
     

Metrics

• Custom Metric Adapter: can be used to get workload’s metrics
• 

In-place resize of Pod Resources

• Still under beta phase (1.33 version)
• To use this: set FEATURE_GATES=InPlacePodVerticalScaling=true
• With this option enabled, the change in resize request won’t kill the pod ```bash apiVersion: apps/v1 kind: Deployment metadata: name: nginx-deployment labels: app: nginx spec: replicas: 3 selector: matchLabels: app: nginx template: metadata: labels: app: nginx spec: containers: - name: nginx image: nginx:1.14.2 ports: - containerPort: 80 resizePolicy: - resourceName: cpu restartPolicy: NotRequired - resourceName: memory restartPolicy: RestartContainer <—–memory can’t be adjusted dynamicaly resources: requests: cpu: “1” memory: “256Mi” limits: cpu: “500m” memory: “512Mi”

```

References

VPA vs HPA