A pod in CrashLoopBackOff keeps its Kubernetes PodDisruptionBudget at zero allowed disruptions and blocks a node drain. Which PDB setting lets the drain evict running pods that aren't Ready?

Answer: unhealthyPodEvictionPolicy: AlwaysAllow

Explanation

With AlwaysAllow, running pods that aren't Ready can be evicted whether or not the budget is met, while healthy pods stay protected. The default, IfHealthyBudget, only allows that when the application still meets its budget, so a crash-looping pod can block a drain indefinitely. The field has been stable since Kubernetes 1.31, and the upstream drain guide recommends AlwaysAllow.