Install failing due to network interface names changing across reboots.  I am using status IP addressing within the install-config.yaml and have to provide the interface name (I've validated with nmstatectl gc <filename>).

I initially got this by booting to a DVD and noting the interface and MAC address - it was ens1f0. When openshift-baremetal-install process ran and installed the OS via Ironic, the interface was named eth3. So I figured RHCOS used a different "predictable" name than RHEL. 

I updated install-config.yaml and re-ran the installation but now the interface was called ens1f0 again.

FYI I'm using Dell PER350 servers all with the Intel X710 NICs installed in the same slot (NIC Slot 1, Port 1).

openshift-baremetal-install 4.12.31
built from commit ead355d7836a8a24613d062bf06ab0a6eef2765c
release image quay.io/openshift-release-dev/ocp-release@sha256:4e0e74e9fdfbaec726654fa883b2018beb548e12ad9db2ffbeb7cf2d6862f7e8
release architecture amd64

Include the following in install-config.yaml:

networkConfig:
  interfaces:
  - name: ens1f0
    type: ethernet
    state: up
    ipv4:
      enabled: true
      address:
      - ip: "10.51.0.33"
        prefix-length: 24
  dns-resolver:
    config:
      server:
       - 10.51.0.201
  routes:
    config:
      - destination: 0.0.0.0/0
        next-hop-address: 10.51.0.1
        next-hop-interface: ens1f0

1. Create clusterconfigs directory
2. Create install-config.yaml in that directory and use the above nmstate config (per controlplane node)
3. Initial RHCOS installation takes place, and from login prompt you can see the interface name changes. Subsequent reboot then fails (possibly because NetworkManager-wait-online.service fails?)

Interface name changes each time the installation is run.

Interface name is consistent (and should also follow the predictable naming, which it doesn't appear to be doing).