cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1742
Views
0
Helpful
19
Replies

Redundant WLC question

ds6123
Level 1
Level 1

Hello,

I’m a little unclear on how WLC redundancy works and can’t quite find the details I’m looking for in the documentation.

Suppose I have 2 WLCs (WLC1 and WLC2…both in the same mobility group) and each has a license for 50 APs. Also, suppose I have 60 access-points in different subnets than the WLC management addresses (yes, I know 10 APs will be out of luck if a WLC actually fails… but that’s a separate issue).

I have a DNS record that points cisco-capwap-controller.domain to the IP address of WLC1.

I believe I read that the APs will learn about other WLCs in the mobility group when the AP joins the WLC.  Does that only work if the AP successfully joins/registers with a WLC and is “primed”?

When the 51st AP tries to join WLC1, will WLC1 ignore it? Or will WLC1 somehow direct the AP to go to WLC2 (since it’s in the same mobility group)?

I’m testing this out, and things seem to work (50 APs on the first WLC and 10 on the second). But I’m not sure if that’s due to me upgrading the Ethernet switch software and causing the APs to move around and giving them a chance to register to WLC1 at one point in time. If I plugged 60 brand new APs (absolutely no previous “primed” connections) in all at the exact same time. Would 10 APs be unable to register to a WLC?

I’m aware that I should probably tell each AP (via the WLC or NCS templates) to use a specific primary, secondary and tertiary controller. But I’m just wondering about the minimum config required.

Any info would be appreciated.

19 Replies 19

George Stefanick
VIP Alumni
VIP Alumni

You are correct with your statement. Once a ap joins the wlc, it will learn about the other controllers in the mob group regardless of "how" you configure the ap, manual prime or other wise.

If your wlc1 fails you should, best practice, have the same WLANs and matching subnets. Ap groups etc on wlc2.

The ap actually selects what controller to go to. If wlc1 is loaded it will join wlc2, because it's in the mob group list. You can be more specific with where the ap joins with ap high availability tab.

You're on the right track ..

Sent from Cisco Technical Support iPhone App

"Satisfaction does not come from knowing the solution, it comes from knowing why." - Rosalind Franklin
___________________________________________________________

Leo Laohoo
Hall of Fame
Hall of Fame

I believe I read that the APs will learn about other WLCs in the mobility group when the AP joins the WLC.  Does that only work if the AP successfully joins/registers with a WLC and is “primed”?

When the 51st AP tries to join WLC1, will WLC1 ignore it? Or will WLC1 somehow direct the AP to go to WLC2 (since it’s in the same mobility group)?

The first fifty will join with WLC1.  The 51st and succeeding WAPs will be in "limbo-land":  They'll be waiting for the WLC to have an opening slot and, since they won't have any configuration, the WAPs will not be servicing any WLAN clients.

If one of the 50 WAPs "pops out" of the controller then any one of the WAPs in "limbo-land" will go in.

grabonlee
Level 4
Level 4

Hi

If your clients will not roam across the WLCs, then there is no point having a mobility group with both WLCs. What I mean is that if a client doessn't roam across 2 APs connected to the 2 WLCs, then it's a waste having each WLC having copies of CAPWAP messages of the other. What you can do for failover is to either specify the Secondary WLC IP name and address on each AP or go to the Wireless Tab ----> Global configuration for all access points and put the details of the backup WLCs. This should be done on both WLCs.

The AP doesn't need mobility groups to join a controller. When a virgin AP boths up, it sends a discovery request via broadcast on its LAN or uses DNS. The WLC checks its configuration to ensure it has capacity to add an AP, if not it doesn't respond to the discovery request.

fbarboza
Level 4
Level 4

Hi,

As reference you can use the follwoing link for deploying the WLC 4400 to get an understanding on how to setup WLC redundancy.

Check under

Controller Redundancy and Access Point Load Balancing

http://www.cisco.com/en/US/docs/wireless/technology/controller/deployment/guide/dep.html#wp1170102

As stated above you are in the correct track.

Thanks everyone for the great answers.  I appreciate it.

Does anyone know how the AP will respond if I create 2 A records for

CISCO-CAPWAP-CONTROLLER (one for each WLC)? Specifically how a brand new, fresh-out-of-the-“eco-pack”-box, AP would respond?

Will it try one, if that fails (ie the WLC refuses the discovery request or doesn’t answer at all), will it try the next A record?

My goal is as little user intervention (on the WLC/WCS/NPS) as possible.

Using DHCP option 43 seems possible but a little impractical (multiple DHCP scopes, multiple DHCP servers, etc.) DNS seems to be the most likely way. I’m just not sure how the AP would respond.

If you use a A record you add the IP of the controller. The AP will join the that controller after which time it will learn about the other controllers .

"Satisfaction does not come from knowing the solution, it comes from knowing why." - Rosalind Franklin
___________________________________________________________

Thanks George,

But here's my problem.  In my example above, I have 2 WLC's with a 50 AP license each.  If I plug in 60 brand new APs, my understanding is 10 will be in limbo-land until they can register with the first WLC.   If that first WLC has exhausted the license, my understanding is it will not tell the AP about the other WLC in the mobility group.  Is this not the case?  So I'm wondering if I have 2 DNS A records (respond with 2 IP addresses in the DNS response) how the AP will behave.  It seems like an obvious thing to me but it's not mentioned anywhere.  Maybe there's a good reason for that, though. 

I see your point. Ive never done it the way you are suggesting with DNS. I have with option43, as you can can add multiple controllers. Another option you can do, as the new <60> aps join you can start moving them over to the second WLC with the high avaibility option. You could even manual prime, but that could be a pain.

"Satisfaction does not come from knowing the solution, it comes from knowing why." - Rosalind Franklin
___________________________________________________________

No you only need one A record so long as the WLC are in the same mobility group.

When the ap sends itsfiscover request the WLC responds with all the WLC in the groups, their ap count and the excess availability.

Steve

Sent from Cisco Technical Support iPhone App

HTH,
Steve

------------------------------------------------------------------------------------------------
Please remember to rate useful posts, and mark questions as answered

Is that right .. I was always under the understanding that it was when the controller took the code that it learned about the mobility group and the WLCs in it ..

Are you going to make me test it this late in the day !

"Satisfaction does not come from knowing the solution, it comes from knowing why." - Rosalind Franklin
___________________________________________________________

Can you George? Your the best! Let us know what your findings are:)

Sent from Cisco Technical Support iPhone App

-Scott
*** Please rate helpful posts ***

HAHAHAH Scott ...

Scott what do you know about Steve's response..

I did find this, but not worth much ..

If the LAP was registered to a WLC in a previous deployment, the LAP maintains the list of WLC IP addresses locally in NVRAM. The stored WLC IP addresses include all of the WLCs that are in previously joined WLC "mobility groups". This is the discovery process:

"Satisfaction does not come from knowing the solution, it comes from knowing why." - Rosalind Franklin
___________________________________________________________

It's at discovery that the AP learns all the WLC in the mobility group. The WLC I supposed to respond with all the WLC in the group then then the AP decides which to join based on greatest excess availability

Per or earlier convo 'should'

Steve

Sent from Cisco Technical Support iPhone App

HTH,
Steve

------------------------------------------------------------------------------------------------
Please remember to rate useful posts, and mark questions as answered

Scott Fella
Hall of Fame
Hall of Fame

That is true. The fastest and only way I clear that is to issue the following:

debug capwap console cli

erase /all nvram:

undebug all

reload

Sent from Cisco Technical Support iPhone App

-Scott
*** Please rate helpful posts ***
Review Cisco Networking products for a $25 gift card