For upgrades and migrations we introduced a check to avoid them when overlapping CIDRs exist - https://github.com/openshift/ovn-kubernetes/pull/2313.

However the installation allows it which ends with the installation broken and no information from anywhere to easily lead to the issue.

For example if I create a install-config with clusternetwork CIDR overlapping with OVNK-I transit switch, everything goes on and at certain point the install will just get stuck and only by accessing one of the masters and seeing the ovnkube-controller logs we can see what is the issue. And this is because I know what I was doing wrong since I configure this on purpose.

Our installation documentation has no mention of this CIDR (only the join subnet) and we can't expect customers to read all other documents prior to install the clusters as we can see in the example below:

https://docs.openshift.com/container-platform/4.15/installing/installing_bare_metal/installing-bare-metal-network-customizations.html

Version-Release number of selected component (if applicable):
OCP 4.14 and 4.15

every time    

    1. Install a cluster with clusternetwork CIDR that overlaps with the OVN internal transit switch subnet
    2.
    3.