Explain HPA vs VPA in Kubernetes.

In Kubernetes, HPA (Horizontal Pod Autoscaler) scales the number of pod replicas based on metrics, while VPA (Vertical Pod Autoscaler) adjusts CPU and memory per pod to optimize workload performance and cost. <a id="definition"></a>

When to Use

• HPA: Best for stateless, bursty workloads like APIs or web servers with unpredictable traffic.
• VPA: Ideal for steady, long-running batch jobs or services with mismatched resource requests.
• Together: Use HPA to handle spikes, VPA to right-size pods.

Example

• HPA: Food ordering app adds replicas during lunch rush.
• VPA: ETL job lowers over-allocated memory requests at night.

Want to master autoscaling decisions for interviews? Check out Grokking System Design Fundamentals, Grokking the Coding Interview, and practice Mock Interviews with ex-FAANG engineers.

Why Is It Important

Using the right autoscaler improves reliability, ensures apps meet SLAs, and reduces cloud cost.

Interview Tips

• Start with: “HPA = replicas, VPA = resources.”
• Mention metrics (CPU, memory, custom).
• Show awareness of safe coexistence and cost implications.

Trade-offs

• HPA: +scales fast, +resilient; −more pods, cold starts.
• VPA: +cost savings, +predictable performance; −pod restarts, slower for sudden bursts.

Pitfalls

• Letting VPA and HPA fight over CPU/memory.
• Ignoring PodDisruptionBudgets—VPA evictions can cause downtime.
• Sticking to defaults instead of tuning thresholds and cooldowns.