Autoscaling Is an Authority System, Not a Capacity System

Short summary

Autoscaling configurations encode governance decisions about what systems can execute without human approval. Most organizations set these once and never revisit them as workloads evolve, creating Scaling Divergence: technically-correct behavior that contradicts operational intent. Real-world consequences range from replica limits hit during traffic spikes to coordination conflicts between HPA and VPA policies.

• •Autoscaling is governance: each parameter encodes a delegation of authority to the execution plane
• •Scaling Divergence occurs when workload characteristics change but scaling policies don't
• •Three failure patterns: max replicas exhausted, min replicas outdated, uncoordinated HPA+VPA

Generated with AI, which can make mistakes.