The Pragmatic Guide to Kubernetes Pod Autoscaling: HPA, VPA, and KEDA

Kubernetes autoscaling is powerful but often misconfigured. This guide covers the three autoscalers: HPA for replica count based on CPU/memory or custom metrics, VPA for right-sizing pod resource requests, and KEDA for event-driven scaling from Kafka, RabbitMQ, Prometheus, and more. Each section includes recommended configurations, common pitfalls, and monitoring strategies using Prometheus metrics. We also discuss the interaction between HPA and Cluster Autoscaler, and when to combine VPA with HPA.