Skip to main content

Module 2: Autoscaling

Overview​

Your Voting App has smart scheduling from Module 1, but every service still runs exactly one replica. What happens during a flash sale when thousands of votes pour in? Static replicas cannot handle traffic spikes. Manual scaling with kubectl scale is reactive, not proactive. You need autoscaling to make the app self-adjusting.

This module teaches horizontal pod autoscaling (HPA) with Metrics Server, vertical pod autoscaling (VPA) recommendations, and event-driven scaling with KEDA. You will configure vote and result services to auto-scale on CPU, then explore scaling the worker based on Redis queue depth.

What You Will Learn​

  • Install and verify Metrics Server in your KIND cluster
  • Configure HPA with CPU targets for vote and result services
  • Generate load and observe autoscaling in real-time
  • Understand VPA recommendations for right-sizing resource requests
  • Set up KEDA for event-driven worker scaling based on Redis queue length

Prerequisites​

  • Module 0 and Module 1 completed
  • KIND cluster running with 3 worker nodes
  • Example Voting App deployed with scheduling rules from Module 1
  • Basic understanding of resource requests and limits

Module Metadata​

Difficulty: Intermediate Estimated Time: 100 minutes

  • Reading: 15 minutes
  • Lab: 60 minutes
  • Quiz: 15 minutes

Ready to make your Voting App respond dynamically to load? Continue to the reading materials.