Two coredns pods was _never_ going to be enough for a cluster this size; a failure was always going to happen at some point. I've seen clusters less than half t

Two coredns pods was _never_ going to be enough for a cluster this size; a failure was always going to happen at some point. I've seen clusters less than half this size kill their coredns deployment with too many requests. With clusters this size you should really be running something like nodelocaldns to cut down on requests to coredns.