In the current code, there is an unnecessary precheck that checks if kolla_internal_vip_address is in the same network as network_interface. In our architecture, we specifically chose to use separate api networks for controllers, network nodes and compute nodes to be able to scale independently. In this scenario, deploy outputs as follows:
TASK [prechecks : Checking if kolla_internal_vip_address is in the same network as network_interface on all nodes] ***
fatal: [10.242.4.14]: FAILED! => {"changed": false, "cmd": ["ip", "-4", "-o", "addr", "show", "dev", "ens3"], "delta": "0:00:00.003956", "end": "2016-07-27 12:29:03.939768", "failed": true, "failed_when_result": true, "rc": 0, "start": "2016-07-27 12:29:03.935812", "stderr": "", "stdout": "2: ens3 inet 10.242.4.14/24 brd 10.242.4.255 scope global ens3\\ valid_lft forever preferred_lft forever", "stdout_lines": ["2: ens3 inet 10.242.4.14/24 brd 10.242.4.255 scope global ens3\\ valid_lft forever preferred_lft forever"], "warnings": []}
...
I believe this check is not necessary and should be removed.
This precheck has been included as part of the below bug fix /bugs.launchpad .net/kolla/ +bug/1566571
https:/
It is mentioned that, as per kolla's requirement, kolla_internal_ vip_address must be in the _same_ network range as network_interface on all node. Without this precheck, many new users are seeing issues and filing invalid bugs.