I installed Openstack with Ubuntu's Autopilot (openstack-install). The bootstrap from the first environment, called "maas" is a VM, that I took from the tools folder insdide the openstack-installer package.
inside that VM are six containers and every one of each has a service (apache2, haproxy, landscape, landscape-msg, postgresql and rabbitmq-server). From the landscape container is "launched" another environment (called "8" in my case), where are all the openstack services.
Everything worked fine until my first reboot. Landscape was unreacheable and the reason was that for juju, landscape has the IP 10.0.3.1, which is the bridge (lxcbr0) IP adress instead of it's eth0 IP address (10.222.221.140 in my case). What was the result? when it tried to connect to landscape, it went to 10.0.3.1, so it went to the VM instead of the container where landscape is running.
I figured out that was a juju-core bug, according to https://bugs.launchpad.net/juju-core/+bug/1416928, so I updated from juju-core 1.20.x to 1.21.3. Reboot and it took the correct IP, as you can see here:
$ JUJU_HOME=~/.cloud-install/ juju status landscape/0 environment: maas machines: "0": agent-state: started agent-version: 220.127.116.11 [...] containers: 0/lxc/2: agent-state: started agent-version: 18.104.22.168 dns-name: 10.222.221.140 [...] services: landscape: [...] units: landscape/0: [...] public-address: 10.222.221.140
The problem is that now the service is not running:
landscape/0: agent-state: error agent-state-info: 'hook failed: "config-changed"' agent-version: 22.214.171.124 machine: 0/lxc/2 public-address: 10.222.221.140
I've tried with
juju resolved --retry landscape/0, but is not working.
Any clues why this is happening?