Kubernetes

Missing HPA stabilization window causes scale-down thrashing

warning
configurationUpdated Mar 25, 2026
How to detect:

HPA without stabilization windows rapidly scales down during brief traffic dips, then immediately scales up again, causing cold starts, cache invalidation, and latency oscillation

Recommended action:

Configure HPA v2 behavior.scaleDown.stabilizationWindowSeconds to 300 seconds to prevent rapid scale-down during spiky traffic. Keeps caches warm and reduces churn