Mirantis Fuel 6.0 HA deployment Failure Wrong IP for NOVA-API on port 8774

Ubuntu 12.04
Neutron with GRE segmentation
Two interfaces
    eth0 for private (192.168.0.0/24 with vlan tag 101 and 192.168.1.0/24 with vlan tag 102)
    eth1 for public (172.16.0.0/24)
Not using Ceph and no additional components installed.

Using Mirantis Fuel 6.0 trying to get a basic HA deployment working one bare-metal with GRE segmentation on Ubuntu 12.04. The Fuel installation uses all default values.

Every deployment fails on the first controller node. Attempts were made with as few as a single controller node and as many as three controller nodes. Each time the results are the same.

Looking at the logs, I can see that port 8774 is failing to be found and causes a time out. The IP address that is being used is on the public network, 172.16.0.0/24, instead of the private 192.168.0.0/24 or 192.168.1.0/24 like all the other services.

The error from the puppet log is
(/Stage[main]/Osnailyfacter::Cluster_ha/Exec[create-m1.micro-flavor]/returns) ERROR (ConnectionError): HTTPConnectionPool(host='172.16.0.3', port=8774): Max retries exceeded with url: /v2/d5092305e5bd4bdab9b0435b85a68267/flavors (Caused by <class 'socket.error'>: [Errno 113] No route to host)

The actual address that has a listener on port 8774 is 192.168.0.5 on the br-mgmt interface.

Network verification passes and regressing to uses multi-node instead of multi-node with HA works fine.

Link to the diagnostic snapshot https://drive.google.com/file/d/0B49gTqY3Dhdmb2phQ29TMURhX1U/view?usp=sharing