1. Proposed title of this feature request

Configure keepalived with unicast instead of multicast when using IPI installation workflow for OpenShift on OpenStack

2. What is the nature and description of the request?

Currently OCP OSP IPI is using keepalived in multicast mode which can lead to VRRP ID clashes if multiple clusters with similar cluster name are deployed on the same L2 network.

The request here is to move OpenStack IPI deployments to a setup where keepalived is using unicast and hence avoiding router id clashes leading to cluster outages, similar to what has been addressed for BM installations and vSphere-based installations (RFE-2634)

3. Why does the customer need this? (List the business requirements here)

Customers are creating numerous OCP clusters in the same L2 subnet on request, for testing and also in an automated way, for CI/CD purposes. This is all done automatically through a process which creates clusters on RHOSP via IPI install method. As cluster naming depends on the project requesting the cluster or starting CI/CD pipeline, cluster names are quite similar. This can lead to clashing router IDs (VRRP) as these are calculated based on cluster name. The implication of this, is that automated tests or user tests (depending on the cluster request) fail because the clash leads to API unavailability.

4. List any affected packages or components.

Bootstrap, control plane, worker and keepalived