Description of problem:

If a cluster is installed as Shared Gateway Mode and after the installation the host default gateway is modified pointing an interface different from br-ex, the default route in the gateway router is not changed causing the traffic from pod network to continue to use the old default gateway. 

Then, if the Gateway Mode is changed from Shared to Local, and then back to Shared the default route in the Gateway Router is completely different causing an unexpected behavior different from the starting point. That may cause for example Core DNS outage or cluster upgrade stuck.  

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

4.20

How reproducible:

100% reproducible

Steps to Reproduce:

1. Install a cluster with multiple interfaces, with Shared Gateway Mode and host default gateway via br-ex

2. Change the host default gateway to use a secondary interface 

3. Verify the default route in the OVN Gateway Router still points the gateway in the network at the br-ex interface.

Actual results:

In case of Gateway mode change, from Shared to Local and then back to Shared, the route is different and pod-to-api communication is unexpectedly broken.

Expected results:

Consistency in the OVN configuration.

Additional info:

This is an issue that was identified in a customer cluster.