But the OpenShift IPI installation on bare metal fails and all nodes show as `localhost` in Red Hat OpenShift Container Platform 4.6

Issue

We are following the IPI installation doc for this deployment:
https://docs.openshift.com/container-platform/4.6/installing/installing_bare_metal_ipi/ipi-install-installation-workflow.html

But the OpenShift IPI installation on bare metal fails and all nodes show as localhost.

The installation starts the automated PXE boot driven by the installer on all servers and the node creation and deployment work, too. But then, the installation fails on a connection refused to reach the cluster API, like this:

DEBUG Still waiting for the Kubernetes API: Get "https://api.cluster.example.com:6443/version?timeout=32s": dial tcp 10.33.200.10:6443: connect: no route to host
INFO API v1.19.0+8d12420 up
INFO Waiting up to 30m0s for bootstrapping to complete...
E0302 15:02:23.408845   61999 reflector.go:307] k8s.io/client-go/tools/watch/informerwatcher.go:146: Failed to watch *v1.ConfigMap: Get "https://api.cluster.example.com:6443/api/v1/namespaces/kube-system/configmaps?allowWatchBookmarks=true&fieldSelector=metadata.name%3Dbootstrap&resourceVersion=7400&timeoutSeconds=370&watch=true": dial tcp 10.33.200.10:6443: connect: connection refused

After going through the steps from https://access.redhat.com/solutions/5551521, it is clear that introspection goes through successfully:

[akaris@supportshell ironic-logs]$ grep 'Introspection finished successfully' ironic-inspector.log 
2021-03-16 15:26:48.641 1 INFO ironic_inspector.process [-] [node: b661395d-9814-4378-8595-d2488a372381 state finished MAC <mac> BMC 10.0.0.19] Introspection finished successfully
2021-03-16 15:27:45.897 1 INFO ironic_inspector.process [-] [node: 5a44be26-b3af-4682-96e5-a1020f262ca1 state finished MAC <mac> BMC 10.0.0.17] Introspection finished successfully
2021-03-16 15:28:03.369 1 INFO ironic_inspector.process [-] [node: dbeef518-9e44-494c-a530-86440e20e05c state finished MAC <mac> BMC 10.0.0.16] Introspection finished successfully

[root@293f3f93f2ac /]# openstack baremetal node list
+--------------------------------------+------+--------------------------------------+-------------+--------------------+-------------+
| UUID                                 | Name | Instance UUID                        | Power State | Provisioning State | Maintenance |
+--------------------------------------+------+--------------------------------------+-------------+--------------------+-------------+
| 5a44be26-b3af-4682-96e5-a1020f262ca1 | master-0 | 6117b9c4-020d-487d-9121-6e987910dce5 | power on    | active             | False       |
| dbeef518-9e44-494c-a530-86440e20e05c | master-1 | 7074dd8b-25d7-47a7-b097-02412ce177b8 | power on    | active             | False       |
| b661395d-9814-4378-8595-d2488a372381 | master-2 | d56a1027-55a4-495f-8fa3-af35a37757f2 | power on    | active             | False       |
+--------------------------------------+------+--------------------------------------+-------------+--------------------+-------------+

However, the installation stalls. When connecting via SSH to the master nodes (core@<IP address>) they all show localhost as the hostname and oc get nodes shows only one node named localhost:

[kni@console ~]$ oc get nodes
NAME        STATUS     ROLES    AGE   VERSION
localhost   NotReady   master   86m   v1.19.0+8d12420