I'm pleased to notice that we have three smallish servers (cloudceph200[1-3]-dev) racked in codfw1dev dedicated to Ceph development. Let's get them up and running and backing the -dev nova and glance services.
Either we can make each node a combined mon and osd node, or we can repurpose labtestpuppetmaster2001 (T258103) as a single mon node. Let's try the former unless applying the two puppet roles in combination causes trouble.
Right now those nodes are on public IPs. In order to replicate eqiad1 we need to create a cluster network for the backplane and then rebuild these nodes on the VM vlan (probably cloud-hosts1-b-codfw).