Is there any chance that the VM's disk was not clean? I know we saw similar failures in MAAS when it would re-use a machine that had been a controller previously, and the disk was not wiped. From the detailed log, it does look like this was a clean install (apt install juju-mongodb3.2 took 1.5min to complete). So that doesn't seem to explain it. If it is a fresh install, then it would be very surprising for the hosted model "(eg 'default')" to already exist. I don't see anything in the above that would cause us to change the name of either "controller" or "default" unless there is config somewhere in ~/.local/share/juju that is defining this cloud with something custom. On Tue, Dec 19, 2017 at 2:01 PM, Kevin Wennemuth <