infra-kubernetes-autoscaling-review

What problem does it solve?

This Skill helps you review and tune Kubernetes autoscaling so workloads scale reliably under real demand without instability or wasted cost.

Core Features & Use Cases

• HPA tuning and diagnostics: Validate min/max replica bounds, metric types (CPU/memory/custom/external), stabilizationWindowSeconds, and ensure resource requests are set to make utilization-based scaling work.
• HPA vs VPA vs KEDA decisioning: Choose horizontal (rate-based), vertical (unknown resource needs), or event-driven (queue/event backlog) autoscaling for the right workload pattern.
• Cluster Autoscaler and availability safeguards: Configure scale-down delays and utilization thresholds, use Pod Disruption Budgets to prevent mass disruption, and optimize behavior for spot nodes.
• Use Case: If an orders API either never scales down or flaps during traffic spikes, use this Skill to diagnose metrics availability, readiness gating, and selectPolicy/behavior settings, then decide whether to switch to KEDA for Kafka lag-based scaling.

Quick Start

Tell the AI: "Review our HPA/VPA/KEDA strategy for the orders-api and orders-consumer, then recommend concrete Kubernetes settings to stop flapping and enable correct scale-down while keeping SLOs."