I'm having an academic debate with a colleague regarding the fail-over behavior amongst RADIUS servers configured on a WLAN controller. My understanding is that unlike the autonomous APs, there is no such thing as a 'dead' timer and that the failover behavior is like this:
1. Primary AAA server available and ALL requests go to this server (unless you manually change the order of primary, secondary, tertiary on a per-WLAN basis to effectively 'load-balance').
2. The primary AAA server is used until it goes down, at which time the secondary AAA server will be used until it cannot be contacted. Once this goes down, the controller will use the tertiary, and then back to the primary.
My argument is that there is no pre-emptive recovery of the AAA server back to the previous one when it comes back online. I have not actually played with this in the lab, but I don't see any parameters in the GUI or the CLI that would allow me to effect the failover behavior. I have also referenced the full CLI guide for all controller releases and have not found anything useful in this situation.