0. Background As part of my day job, I help to manage a large Openstack cluster with a large number of virtual instances running on physical infrastructure. Things often go wrong, and sometimes you have to get creative to solve them. Today, someone reported that a particular instance had failed catastrophically (due to a layer 8 malfunction), and could not be recovered. We decided it would be easier to rebuild a new instance rather than try and fix the instance (it was well and truly borked).