Kubernetes Cost Optimization and Autoscaling

🛠 Lab from the Platform Engineering Bootcamp. Used in Week 21. Bootcamp landing page: https://academy.tekanaid.com/bootcamps/platform-engineering-bootcamp Parent course(s):

• Week 21: Platform Hardening & Production Readiness: Capstone Part 2 (slug: platform-hardening-production-readiness)

Give your Internal Developer Platform real cost visibility and real elasticity. Working against the real TaskFlow backend on a Minikube cluster, you'll deploy Kubecost and query its live cost-allocation API to build per-team showback and chargeback reports, pull real Vertical Pod Autoscaler recommendations to right-size resource requests, drive a genuine Horizontal Pod Autoscaler scale-up under live load, hand autoscaling over to KEDA for event-driven queue-based scaling, and enforce a FinOps budget with a real, live-tested ResourceQuota.

Every check in this lab verifies a real live effect, not just object existence: real dollar figures out of Kubecost's allocation API, a real VPA recommendation computed from real usage samples, a real HPA scale-up under real load, a real KEDA-driven scale-up triggered by real queue depth, and a real ResourceQuota rejection when a deployment tries to exceed its budget.

Key Learning Objectives:

• Deploy Kubecost and query its live cost-allocation API for real per-namespace and per-team cost data
• Build showback and chargeback reports from real Kubecost cost data
• Read real Vertical Pod Autoscaler recommendations to right-size CPU and memory requests
• Configure a Horizontal Pod Autoscaler and trigger a real CPU-driven scale-up
• Install KEDA and drive event-driven, queue-depth-based autoscaling
• Enforce namespace cost budgets with ResourceQuota and verify real admission-time rejection