Wireless Access

We are running three 7220 controllers in a L2 cluster on 8.5.0.2. I noticed that the client load balancing is all off and almost all clients are on one controller. There are a few hundred clients on the other two controllers, but in general, the majority are on one controller (the cluster leader).

Redundancy Mode On
Active Client Rebalance Threshold = 50%
Standby Client Rebalance Threshold = 75%
Unbalance Threshold = 5%
AP Load Balancing: Enabled
Active AP Rebalance Threshold = 20%
Active AP Unbalance Threshold = 5%
Active AP Rebalance AP Count = 50
Active AP Rebalance Timer = 1 minutes
Cluster Info Table
------------------
Type IPv4 Address Priority Connection-Type STATUS
---- --------------- -------- --------------- ------
peer x.x.x.1 128 L2-Connected CONNECTED (Leader, last HBT_RSP 82ms ago, RTD = 0.476 ms)
self x.x.x.2 128 N/A CONNECTED (Member)
peer x.x.x.3 128 L2-Connected CONNECTED (Member, last HBT_RSP 77ms ago, RTD = 0.403 ms)

x.x.x.1:

ipv4 wireless Users: 7434
mac auth: 224
dot1x: 7186
captiveportal: 12
other: 12

x.x.x.2:

ipv4 wireless Users: 223
mac auth: 178
dot1x: 35
captiveportal: 5
other: 5

x.x.x.3:

ipv4 wireless Users: 196
mac auth: 139
dot1x: 41
captiveportal: 9
other: 7

What's going on here? Anyone else having the same issue?

What was the last infrastructure change that you made?

What is the output of "show lc-cluster load distribution client"?

The last change we made was months ago when we added an additional controller. Early in the morning with not a lot of clients, here is the output:

peer x.x.x.1 48 1639
self x.x.x.2 1673 47
peer x.x.x.3 40 19

Active Client Rebalance Threshold = 50%
Standby Client Rebalance Threshold = 75%
Unbalance Threshold = 5%



peer x.x.x.1 48 1639
self x.x.x.2 1673 47
peer x.x.x.3 40 19

The platform limit of a 7220 is 24,576

For Active user load balancing to happen, two things must be true:

1. Active Client Rebalance Threshold = 50% - One controller must have 50% of its platform limit or greater

2.  Unbalance Threshold = 5% - The difference in users between two controllers  the most heavily loaded controller and the least loaded controller must be greater than 5%

Since your Active Client Rebalance Threshold is not being met, there is no active user load balancing.  

So I'm not sure I understand why there is no rebalancing. The difference between two controllers needs to be greater than 5%? One controller has the bulk of the load, why does that not factor in as one of the two controllers?

I mean right now we have about 5k clients on and the load balance is the same except the cluster leader now has 5k instead of 1k. The difference between .1 and .2 is way more than 5% as is the difference between .2. and .3. Where there isn't a difference more than 5 percent is between .1 and .3.

So I guess I'm not sure how that unbalance threshold is calculated.

For Active user load balancing to happen, two things must be true:

Oh okay.

Do you think it's worthwhile to adjust the load balancing parameters to get a more even client balance? Or can that cause more issues?

You could certainly try that.

Thank you very much for the information and assistance. Appreciated.