Perhaps I'm missing the point of your post......
But we need to clarify what you are asking for.
From the few posts here, I understand that Client 1 at Site A gets an IP address from site A. When Client 1 roams to Site B, they continue to have the IP address from Site A. You don't want this?
So what is happening:
Layer 3 mobility.
When you roam from Site A to Site B, you should be anchored back to your original controller and should function with your IP address. So from client perspective, absolutely nothing should be wrong.
Now, are you trying to say that Clients that Roamed from Site A to Site B, cannot talk to Site B resources with the Site A address? That would be the only reason I can think of where you'd "not work".
If you absolutely don't want Layer 3 roaming, then you have two options:
1) Do not allow your WLCs to be mobility aware. If they do not have each other in the mobility domain, no mobility will take place and your client will stay Layer 2.
2) (theory) Change your Virtual IP address between the two controllers. Mobility Handoff is rejected if the Virtual IP does not match. So when your client is handed off layer 3, the Site A WLC will not accept it and will remove the client instead of anchor.
The problem with that above however is that you need a client that is stupid enough to re-DHCP as soon as it roams....
If you have IP 10.10.1.10, and you roam to a new stie, with no L3 roaming, you're going to need a client that will either always re-dhcp, or will quickly learn 10.10.1.10 is not valid.....
Long Story short:
L3 Mobility is there for a reason and there is no "disable" button. Either allow your network to function with L3 clients, or configure it so L3 mobility will not happen.
