Description of problem:
During the installation of IPI cluster on Vsphere the API VIP is moved to a master node that has not any kube-apiserver running, 
and is not automatically moved back to the bootstrap node that is the unique node with a running kube-apiserver at that moment.
The effect is the installation blocked and not able to continue.

Version-Release number of selected component (if applicable):
4.7.12 was used by the customer reporting the issue.

How reproducible:
Not sure but on the customer side happened 3/3 times.

Steps to Reproduce:
1.
2.
3.

Actual results:
The keepalived on one of the master node became MASTER for the API VIP without having kube-apiserver running.

Expected results:
The keepalived shouldn't move the VIP at that stage or should be able to detect the VIP not working on the master and move it back to the bootstrap node.

Additional info:
The workaround found at this moment is to manually restart the master node owning the VIP to force the VIP 
to be assigned back to the bootstrap node and allow the installation to finish.