Autoscaling

The application can be scaled manually by setting the number of replicas in the configuration. Additionally, the Helm chart includes an autoscaling mechanism (HPA - Horizontal Pod Autoscaler) that dynamically adjusts the number of replicas based on workload. HPA configuration allows defining:

• Minimum and maximum number of replicas (minReplicas, maxReplicas)
• Target CPU and memory utilization (targetCPU, targetMemory)
• Scaling behavior, such as stabilization windows and scaling policies Example HPA configuration in the Helm chart:

    hpa:
        enabled: true
        minReplicas: 2
        maxReplicas: 10
        targetCPU: 80
        targetMemory: 75

• enabled – Enables or disables autoscaling.
• minReplicas – Minimum number of replicas to ensure the application doesn’t scale below this limit.
• maxReplicas – Maximum number of replicas allowed, preventing over-scaling.
• targetCPU – CPU utilization percentage that triggers scaling up/down.
• targetMemory – Memory utilization percentage that triggers scaling up/down.

Additionally, HPA configuration allows defining scaling behavior for both scaling up (scaleUp) and scaling down (scaleDown).

behavior:
  scaleUp:
    stabilizationWindowSeconds: 120  # Time (in seconds) for which past recommendations are considered before scaling up
    selectPolicy: Max                # Policy selection method ("Max", "Min", "Disabled")
    policies: []                      # Scaling policies for scale up
 
  scaleDown:
    stabilizationWindowSeconds: 300   # Time (in seconds) for which past recommendations are considered before scaling down
    selectPolicy: Max                 # Policy selection method ("Max", "Min", "Disabled")
    policies:
      - type: Pods                    # Policy type (Pods - number of replicas, Percent - percentage change)
        value: 1                      # Number of replicas to add/remove
        periodSeconds: 300            # Minimum time interval between scaling actions

If the application has specific scaling requirements, a custom HPA definition can be added to the chart by modifying values.yaml or creating a separate Kubernetes HPA manifest.