I have the following setup: Ubuntu 14.04 LTS Juju 1.22.1-trusty-amd64 Maas 1.5.4+bzr2294-0ubuntu1.3
MAAS managed cluster with all physical hosts attached to the following networks (each is on its own vlan):
10.92.109.0/24 - MAAS managed network
10.92.104.0/24 - Internal Data Network
10.92.105.0/24 - Internal 'service' network
Maas is aware of the latter 2 but does not manage them directly (they have their own dhcp). Maas provides dhcp and DNS for the 109 network.
I have created an environments.yaml file as follows:
environments: maas: type: maas maas-server: 'http://<MY MAAS SERVER>/MAAS' maas-oauth: <SECRET> admin-secret: '<ANOTHER SECRET>' default-series: 'trusty' bootstrap-timeout: 1800
and bootstrapped it as follows:
juju bootstrap -e maas --to node01.<MY MAAS MANAGED NETWORK> juju sync-tools -e maas juju deploy juju-gui --to "0"
and then some add-machine commands.
This works as expected but when I look at the agent config for the newly added machines I see:
apiaddresses: - 10.92.104.101:17070
Rather than either the FQDN of the maas network home of the host, or alternatively the host's 10.92.109 address.
While this still works for my physical machines it means that management traffic is passing over the 'wrong' network, and means that when I try to deploy a service to an LXC container it doesn't work, as the LXC container is correctly bridged onto the MAAS managed network via the juju-br0 interface.
Specifically the LXC container can't download the juju-tools archive nor would it be able to connect to the api-address it's been given by cloud-init.
Running juju api-info I get:
user: admin environ-uuid: <UUID> state-servers: - 10.92.109.101:17070 - 10.92.104.101:17070 - 10.92.105.101:17070 ca-cert: |
While juju api-endpoints gives 10.92.109.101:17070