High availability deployment Wireless Lan Controller

[ariask93]

Hey guys,

I'm bringing up this topic because we are discussing the best option for a second stack of controllers.

Right now, we have a stack of WLC 9800-40 controllers working fine with around 500 APs registered. This stack is located in "Zone A." However, we want to deploy another stack of 9800-40 controllers in "Zone B," a couple of kilometers away.

The problem is that we want to maintain the same addressing for users. Currently, all the gateways are in the distribution layer of the controller. One option to achieve this is using HSRP, but it is risky and requires a lot of effort to work properly.

The second option is to deploy a separate controller with different addressing and use HA Mobility between both controllers.

In case of failure, all APs would need to rejoin the new controller.

The main goal for both solutions is to minimize downtime as much as possible when switching from one controller to another.

What do you think is the best option? Have you implemented something similar before?

[Scott Fella]

You really don't want to span your vlans, It almost seems that you should just keep the two sites with its own controllers and not try to combine them.  The only other way is using FlexConnect local switching so that devices obtain dhcp from that local subnet. Keep it simple or else troubleshooting and opening a TAC case will be very difficult.

[ariask93]

That's what I want keep it simple.

thanks for your advice!

[Leo Laohoo]

How critical is the WiFi network? 

If it is "that" critical, do not use HA SSO and go with N+1 and reboot the controllers once every 4 to 6 months.

[Scott Fella]

100% @Leo Laohoo 

[Leo Laohoo]

Thanks, @Scott Fella. 

[ariask93]

it's critical but... we haven't face any blackout of the wifi in 15 years, no one site have lost the energy connection... so.. I don't think an scenario like could happen.

N+1 is a good idea.

Thanks Leo for your advice.

[Leo Laohoo]

Read this:  Cisco Catalyst 9800 Series Configuration Best Practices

Read it well.  There is a "hidden message" in all that. 

[Rich R]

This is very similar to the design we use to achieve 99.999% availability.  We have a pair of HA-SSO WLCs in each data centre (opposite sides of the country) and N+1 with mobility configured for the APs between the DCs. We split the client IP range between the 2 so we have 2 contiguous ranges allocated to the 2 DCs.  And each of those IP pools is split between the DHCP servers in each DC so DHCP A can serve WLC B and vice versa.  This means it protects against WLC failure, DC failure (this happened to us once when our power team had a "disaster" during UPS maintenance) and DHCP server failure.  APs are configured with pri/sec WLC (in the AP HA settings) and we set "capwap timers primary-discovery-timeout 600" so APs will take around 10 minutes to switch back to their primary when it recovers (ensuring it's ready to take the load by that time).  They fail over to backup much quicker if the primary fails.  Agree with @Scott Fella to avoid spanning VLANs between DCs - keep them separate and route between them.