You should install on the shared node. What do you mean by how will the Clients now? Will it not be the same dns name or ip?
in a cluster, you have to a Role. a Role defines 2 major things.
1. what is the defined process that is actually clustered. Ie, a service, an app, a process, etc
2. what is the virtual IP for the above service or app or process.
without criteria number 1, when a fail over occurs, how will the new primary node of the cluster know what process or service to start?
without criteria number 2, the client or web URL potentially will be pointing to an inactive secondary.
Now, i could get around criteria 2 with a load balancer, however, in order to set up a load balancer, it also has to listen to a port or process to know when to activate node 1 or node 2.
I could also ask a simpler question; when i failover the cluster, how does the new primary node know to start the VC service?