Vertical Pod Autoscaler was designed to help teams rightsize workloads automatically, but in production, its architectural limitations create more risk than value.
In Nic’s previous video, “The VPA Problem,” we explained why VPA fails under real-world conditions. In this follow-up, we focus on what effective rightsizing actually looks like at the implementation level.
We break down how VPA’s reliance on historical data, delayed recommendations, and eviction-based updates forces teams to disable automation entirely, especially for stateful and high-availability workloads.
You’ll then see how real-time, context-aware rightsizing operates in practice, adjusting resources as load changes without restarting pods or violating availability constraints.
If you’re responsible for optimizing Kubernetes resources in production, this video shows what “done right” actually means.