Consistent network device naming in latest train overcloud image builds is not working

[Wojciech]

Description
===========
Building overcloud images from latest train tree (centos7-train/5d/ca/5dcaa10e894438fbc3b0cd6ba48a2407081d8613_ed45750d) builds overcloud images with disabled consistent network device naming, witch in consequence track to overcloud deployment failures on baremetal servers with more than one Ethernet vendor inside
os-net-config errors deployment as interace numbers change on every deployment (observed on both general nicX, and ethX inputs in nic-configs)

Steps to reproduce
==================
Deploy overcloud (train) on centos7, by building first images according to documentation
https://docs.openstack.org/project-deploy-guide/tripleo-docs/latest/deployment/install_overcloud.html

Expected result
===============
Sucessfull overcloud deployment

Actual result
=============
Error on os-net-config part of deployment caused by non found interface, or wrong numbering (loss contact to hv) on random nodes

to resolve this i had to change /etc/default/grub on overcloud images to not use "net.ifnames=0"

Regards
WS