#Start unconfigured interfaces as down.
1. Currently, one of these has to happen to solve for an interface that automatically ups itself and tries to DHCP an address.:

1. Don't cable extra interfaces until after the cluster is built
2. Disable switch port of extra interfaces that are cabled, then re-enable after cluster is built
3. Disable extra interfaces in bios if possible, then re-enable after the cluster is built
4. Get MAC address of extra interfaces on each host and build an nmstateconfig to explicitly bring those interfaces down for each host

Options 1-3 aren't very good solutions for any automated build and most enterprises operate under the least access principle, so the OCP team should never have switch access.

Option 4 requires custom automation if configuring more than a small number of hosts with few interfaces each.  

The extra nics (for future storage) should be down by default to allow for dozens of clusters to be configured.