k8s offers some of the same benefits but not to the same extent because the restarting and state shedding happens at the level of the a container on k8s. On the

k8s offers some of the same benefits but not to the same extent because the restarting and state shedding happens at the level of the a container on k8s. On the Erlang VM it happens on a per-process (what Erlang calls coroutines/green-threads/fibers), so rather than losing all state and all requests currently happening in a single instance, you only lose the 1 of potentially millions of processes in your program. This means you can handle failure much more gracefully, and within nanoseconds rather than seconds. I like to say that Erlang's supervisors are to k8s what k8s is to datacenter failover. Similar things, different level of abstraction. And yes it's common to use Erlang with k8s, though there is a little redunancy. It's still good for all the other k8s features.