Currently, RHOSO18 can deploy on several OCP configuration.

1. 3 node compact cluster(collocating master and worker role in a node).
2. Dedicated 3x master and 3x worker nodes(master and worker are separated)

The second configuration has a problem described at OSPRH-10442.

To avoid the asymmetric routing during deployment, one possible solution is adding a worker node which doesn't have NICs for OpenStack networks called here as OSP specific worker nodes.

However, all pods except httpd of OpenStackProvisionServer are located in worker nodes which have NICs for OpenStack networks.
We need to add nodeselector to all pods except OpenStackProvisionServer for placing them on the OSP specific worker nodes.