Hacker News Reader

What happens if a node fails?

As long as you have multiple replicas of the pods they keep serving traffic.

So no work rescheduling? That is a significant divergence from k8s and means you cannot "replace" it with this