HPA

HPA changes replica count from metrics such as CPU, memory, or custom signals (queue depth, RPS). It absorbs variable load when requests, metrics, and readiness are configured correctly.

					apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: api
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api
  minReplicas: 2
  maxReplicas: 20
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70
				

CPU-based scaling needs resource `requests`. Custom metrics need metrics adapters and meaningful business signals.

On interviews: HPA with readiness and cold start latency; database bottlenecks; why autoscaling cannot fix inefficient code alone.

Common pitfalls: missing requests make CPU scaling unreliable; scaling app tier overloads databases or third parties.

The trade-off is elastic capacity versus downstream saturation and cost.

Checklist:

• Set resource requests on workloads.
• Scale on meaningful metrics.
• Check downstream capacity.
• Pair HPA with cluster autoscaling thoughtfully.