PAWS currently has a network entry that looks like:
paws.wmcloud.org -> 185.15.56.57 -> paws-k8s-haproxy-[12] -> paws-k8s-ingress-[34] OR paws-prometheus-1 -> k8s cluster, typically a worker
T319366 gives us some reasons to rethink the parts that are not in the k8s cluster themselves. At first glance both the haproxy and ingress nodes are redundant to the cluster itself, as, ultimately, the cluster will handle redirects through its internal networking. It would appear that if we were to absorb prometheus into the cluster itself that both the ingress and the haproxy systems could be dropped.