Comment 5 for bug 1631864

I've found the same issue with cs:xenial/rabbitmq-server-54 on MAAS 2.0: when rabbitmq-server is deployed to a LXD container with multiple NICs, I can see in the unit log "getting local hostname" and trying to set rabbit@eth2 (which fails). Trying to reverse-resolve the private address 10.100.0.110, returns 'eth2.juju-xyz-machine-1-lxd-2.maas', which is typical with MAAS 2.0. So I think using "local hostname" is wrong, it should use the FQDN or at least that part of it which ends with $(hostname) - in my case "eth2.juju-xyz-machine-1-lxd-2.".

Note that, even though --bind 'amqp=internal-api' was used during deployment, the charm still tries to use the unit-get private-address and access-network (not set) and does not call network-get amqp --primary-address