Description of problem:

Each node has 2 NICs (eth0 and eth1), each of them standing on different networks:

eth0 -> 192.168.10.x/24
eth1 -> 10.10.0.x/24

This, caused the nodes to be registered with both IPs on the cluster:

e.g.:

"name": "ocp-node1",

                "addresses": [
                    {
                        "type": "InternalIP",
                        "address": "10.10.0.7"
                    },
                    {
                        "type": "InternalIP",
                        "address": "192.168.10.11"
                    }
                ], 

Randomly, the hostIP is switched between these IPs:



            "host": "node1",
            "hostIP": "192.168.10.7",
            "subnet": "10.130.0.0/23"

            "host": "node2",
            "hostIP": "192.168.10.4",
            "subnet": "10.131.0.0/23"

            "host": "node3",
            "hostIP": "192.168.10.9",
            "subnet": "10.129.0.0/23"

            "host": "node4",
            "hostIP": "10.10.0.7",
            "subnet": "10.128.0.0/23"

            "host": "node5",
            "hostIP": "192.168.10.10",
            "subnet": "10.130.0.0/23"


Version-Release number of selected component (if applicable):
OCP 3.4.1.2
Openstack

How reproducible:
Almost each time a node is rebooted, it ends up with with a different IP.

Steps to Reproduce:
1. Will add in a comment.
2.
3.

Actual results:
Host IP is switching between eth's IP

Expected results:
Node got registered always with the same Host IP

Additional info:

This bug was originally reported in BZ 1410128, and fixed with errata https://access.redhat.com/errata/RHBA-2017:0218