Description of problem:
In the Sphinx docs for installing instack-undercloud, we list a network topology where nodes boot off of eth1. For many kinds of servers, this will either not work, or will only work for one boot. Many servers can only PXE boot off of eth0.

Version-Release number of selected component (if applicable):
All

How reproducible:
100%

Steps to Reproduce:
1. Install instack-undercloud on undercloud hosts
2. Set overcloud hosts to boot off of eth1
3. Register/discover/deploy overcloud

Actual results:
On some hardware, such as HP ProLiant DL320e Gen8 v2, you can set the system to boot off of eth1, but the system will set itself back to eth0 on next boot. Some systems do not even allow booting from a NIC other than the first (eth0). This means that either discovery will succeed and deployment will fail, or discovery won't even complete.

Expected results:
The topology in the documentation should be workable for all deployments.

Additional info:
The configuration that works for me is to have the undercloud set to use eth0 as the public/uplink, and use eth1 as the PXE boot NIC. The overcloud hosts use eth0 as the boot NIC, and eth1 is available (for a provider external/public link).