Wireless Access

Cluster appears good, vrrp is good, all things seem normal.  Can't get ap's to balance.  Vlan-probe good.     

 

Simple config, how can i force the ap's to balance.  2 7220s 450ish TOTAL ap's.

I'm missing something.

 

Hi

By default AP LB is not enabled on the cluster. You can enable it under system-profiles-classic cluster.

With two MC’s all AP’s will have an tunnel to both MC’s anyway so AP LB should not make a big difference. Client traffic is load balanced, that’s the most important.

AP LB is enabled, sorry i should have clarified that.   I would have expected that I would see half of my aps on the active, and half on the standby.   Otherwise why call it Load Balancing?

The cluster manager balances the load periodically when the total load exceeds the minimum load threshold and the difference between the max and min loaded controller is more than the unbalance threshold.

 

Even with AP load balancing enabled, it can take quite awhile to load balance access points.  It is meant to be gradual to avoid mass disruption.  Since clients might not even be on the controllers that the access points are on, it might have less of an effect than it seems.

 

I'm a little confused on load balancing.  Things seem somewhat inconsistent.  (note that I'm running a pair of 7240's with 1,243 ap's connected)

 

When I look at this document: https://www.arubanetworks.com/techdocs/ArubaOS_81_Web_Help/Content/ArubaFrameStyles/Cluster/Load%20Balancing.htm 

 

This document seems to suggest that client load balancing shouldn't really start until 8,000 clients are connected to the controller however I'm seeing clients appear to load balance after ~ 1,000 clients are connected.

 

However, AP load balancing seems to follow a different model. I was told today that the 7240's are good for up to ~ 2,048 AP's so 1,240 ap's would be a little past 50%, however, I'm not seeing automatic ap load balancing.  I can manually migrate (apmove) ap's to the second controller in the cluster and they migrate, and appear to still carry clients.  The ap's show that they're connected to both controllers. 

 

My cluster show's that it's L2-connected... 

 

 

Is what I'm seeing "expected" behavior?

 

 

Hi

Pre 8.3 AP loadbalancing isn’t on by default. You need to switch it on manually AFAIK.

You can do that on the CLI or under the all profile settings

I understand that by default AP load balancing isn't enabled.

 

My cluster is defined thus:

(MD2) #show lc-cluster group-membership

Cluster Enabled, Profile Name = "MD-Cluster"
Redundancy Mode On
Active Client Rebalance Threshold = 50%
Standby Client Rebalance Threshold = 30%
Unbalance Threshold = 5%
AP Load Balancing: Enabled
Active AP Rebalance Threshold = 50%
Active AP Unbalance Threshold = 5%
Active AP Rebalance AP Count = 10
Active AP Rebalance Timer = 5 minutes
Cluster Info Table
------------------
Type IPv4 Address Priority Connection-Type STATUS
---- --------------- -------- --------------- ------
peer 10.0.0.1 128 L2-Connected CONNECTED (Member, last HBT_RSP 100ms ago, RTD = 0.267 ms)
self 10.0.0.2 128 N/A CONNECTED (Leader)

 

I see that clients do load balance:

 

(MD2) #show lc-cluster load distribution client

Cluster Load Distribution for Clients
-------------------------------------
Type IPv4 Address Active Clients Standby Clients
---- --------------- -------------- ---------------
peer 10.0.0.1 65 848
self 10.0.0.2 852 65
Total: Active Clients 917 Standby Clients 913

 

But at this point, I've deliberately moved some AP's to the second controller because they were not auto balancing...

 

(MD2) #show lc-cluster load distribution ap

Cluster Load Distribution for APs
---------------------------------
Type IPv4 Address Active APs Standby APs
---- --------------- -------------- ---------------
peer 10.0.0.1 474 769
self 10.0.0.2 769 474
Total: Active APs 1243 Standby APs 1243

 

I've seen some examples where auto balancing was happening with less than 100 AP's... but perhaps I shouldn't expect that with a pair of 7240's running as the MD's for this environment.