Question

WiNG controller cluster with extended VLAN and DHCP server on cluster

  • 3 July 2019
  • 4 replies
  • 293 views

We have a cluster of 3 RFS7000 controllers.

One of our SSIDs uses an extended VLAN, i.e. tunnelled to the controller.

The controller is also a DHCP server for clients on the SSID attached to this VLAN, therefore the controller needs an SVI (ip address) valid on that VLAN as well.

Now it seems like a no-brainer that each controller in the cluster would need a different IP address on that VLAN, but is this actually the case? This is how it is set up at the moment, but can anyone confirm that this is definitely correct, or does the failover take care of the same IP address on each controller?

One of the effects of each controller having a different IP address on the VLAN is that the DHCP server is started on all 3 controllers; I presume, however, that there is a method that they use to ensure that they don't issue duplicate IP addresses, although I don't know what that method would be.

4 replies

Userlevel 5
Chris, based on you indicating that you have 3 RFS controllers in a cluster, I'm guessing that we're talking about WiNG-4 here. Yes?
If so, I can't help you with this one, but I'll see what I can find out. Finding answers to this will require someone who has experience with the older OS.
Hi Chris,

No, this is WiNG 5.8.2.0. I'm not even sure that it's a supported configuration to have 3 controllers in a cluster, but that's what I've inherited.

Master priorities are assigned to each one, with the designated primary having the highest, then the secondary, then the tertiary.
Userlevel 3
Hi Chris,

I would suggest (Assuming controller is also your default gateway)using VRRP and DHCP run criteria.
For VRRP provided system reference guide. See page 714.

Note:
Suggest setting establishment criteria as cluster master for both DHCP and VRRP.
Userlevel 5
Chris, okay...so this is WiNG-5 then. In that case, WiNG-5 controller clusters are limited to two controllers (matching), unlike WiNG-4 which supported more controllers for non-related reasons not discussed here.

Is your cluster setup in Active/Active or Active Standby mode?
Certainly, if running in Active/Standby mode, as you mentioned, if a controller is running a DHCP service, it will require an SVI for that VLAN.
But even if running in Active/Active mode, you would still want this setup since a failure of the controller that has the SVI and offering DHCP service might be the one that fails.

As far as the fail-over action - if the cluster is configured in Active/Standby mode, the Standby controller does not automatically inherit the IP address of the Primary. As Daren mentioned, VRRP is supported if you want that sort of topology.

As far as the different DHCP services operating on more than 1 controller and handing out dupe leases- yes, this can happen if you are running Active/Active controller cluster. *CAN* happen.
Though if running Active/Standby, which is the RECOMMENDED method of configuring a cluster anyway, this isn't possible since only 1 controller is active at any given time anyway.

But, just to be thorough here, if you ARE running in Active/Active mode (need to for whatever reason) then you can avoid having both DHCP service operating simultaneously by configuring things so that the DHCP service only operates on the controller that is the current RF Domain Manager....because only 1 of the two controllers will be the active RF-Domain Manager at any given time.

Reply