Why is the PreferNoSchedule taint rarely used in real-world Kubernetes clusters, and what are the operational drawbacks compared to NoSchedule and NoExecute
I’m not sure how strong the use case is. The one concrete example I could find suggested that if you add nodes to your K8s cluster using spot pricing, you might want to prioritize those nodes to more urgent pods, and then allow some other pods to use the excess capacity by using an optional toleration like PreferNoSchedule to sneak in. But beyond that, I couldn’t find a good case for using that parameter at all.