though I guess at some point we should put together a playbook for setting up a large number of nodes...