Extending WLC HA through L2 over the Network.

[Muhammed Adnan]

Hi,

I have a design requirement wherein I cannot connect the redundancy ports back to back on Cisco 8540 WLC.

The requirement is to have the Primary and Secondary WLC in different datacenters. Thus need to connect them over the network over L2 wherein each WLC is terminated on the L2 of ACI service leaf. 

Is my requirement achievable?

Can find the below expert from 8.4 HA config guide. Do we have some detail documentation on same of how the facilitate the needful along with configuration example?

In a management network, you can also use the redundancy management interface (RMI) that is mapped to the redundancy port or RMI VLAN, or both, where the RMI is the same as the management VLAN. Note that the RMI and the redundancy port

should be in two separate Layer2 VLANs, which is a mandatory configuration.

[pwwiddicombe]

The trick is to have a simplistic L2 path between the 2 controllers.  The RP port is incapable of routing; and you also have to make sure you don't have any variable latency on that path (or you may find the HA partner will switch over unexpectedly). 

I suppose in theory it should work with a good quality GRE tunnel.  but be aware that if the tunnel goes down or has interruptions, HA may misbehave.

[Muhammed Adnan]

Thanks Scott and Combe for your valuable responses. 

I need to have mandatorily the AP and client SSO to meet my specification. 

So based on your responses appears that my requirement is possible provided I have my management and dynamic interfaces on the HA pairs on the same subnet, will this enough be fine to meet my criterion?

[Scott Fella]

As long as all the vlans span across and that you don't have less than 100ms latency from the RP's you should be fine.

-Scott 

*** Please rate helpful posts *** 

[Scott Fella]

Just to add. If you plan on using SSO, your interfaces also have to belong on the same subnet. So it's not just the management, alll your dynamic interfaces. 

If this isn't part of your design, then N+1 is what you need to do. Or if you are doing FlexConnect local switching, you can use SSO and just extend the management over.  FlexConnect central switching will require the dynamic interfaces on the same subnet so this design will not work unless you extend the L2 over. 

-Scott 

*** Please rate helpful posts ***