Then there's a lot less reason to run two servers. You still have one central point of failure: the SQL server.

The sql server is actually a mysql master/master replication cluster with one virtual IP address I pointed the servers to. I think this solution avoids s.p.o.f., isn't it?

If you're insistent on running just one SQL server, you don't need to do anything on the RADIUS side for IP pools. Just point both RADIUS servers to the same SQL DB and tables, and the SQL server will sort it out.

Hmmmm... But ip pools are managed through local files on each radius server, the sql backend stores sessions but not ip assignement. Do I miss something? Thanks, Francesco.