CUBE HA setup for new 8200 series router.

[simon hester]

Hi 
Hoping this is the right place for this question. 

I'm working to replace some older 3900 series routers that are acting as CUBE's in a customer CUCM setup. 

The existing CUBE's used HSRP as the HA method, installed before my time but i believe used the following guide Cisco Unified Border Element Configuration Guide Through Cisco IOS XE 17.5 - High Availability on Cisco Integrated Services Routers (ISR-G2) [Cisco Unified Border Element] - Cisco

For the new CUBE's 8200 series I came across the following guide that is very different Cisco Unified Border Element Configuration Guide Through Cisco IOS XE 17.5 - High Availability on Cisco 4000 Series ISR and Cisco Catalyst 8000 Series Edge Platforms [Cisco Unified Border Element] - Cisco

My questions are
Can the new 8200 series CUBE's use the HSRP method?
What is the best HA method HSRP or the one outlined in the guide here which is the one above for 8200 series routers?

Thanks in advance. 

[b.winter · Accepted solution]

You should use the Box-to-Box Redundancy explained in the second link.
I don't even know if HSRP is supported for the CUBE functionality on 4k and 8k router. But this should probably be stated in the second link too.

[simon hester]

Hi 

Thanks for the comment. The HSRP was all done before my time. It does work and there is some magic happening (that i have never been able to quite explain ) that syncs the sessions as if a Cube dies midway through a call the call continues on the other Cube. We also see session data on both cubes. But like i say I've never been able to full understand it all. 

I'll have a look at the Box to Box Redundancy in more detail. 

[Nithin Eluvathingal]

You need to go with the BOX to BOX redundancy. you must have CUBE  A-FLEX-ENH-CUBE license

[simon hester]

Hi thanks for the comments 
Would it change anything given the 2 cubes are split across 2 data centre's. The data centres are physically separate and for redundancy they ideally need to host a CUBE in each data centre.

Thanks

[b.winter]

Not really.