[Q] Maximum Number of Concurrent Connections to 1142 LAP

Question

We are having random issues of users not being able to connect to our wireless network consistently. The users will have successfully accessed the network previously but then will have difficulty associating to the network. After a period of time, the association appears successful again. My first thoughts were that there was a restriction on the number of clients that could associate to a given AP at any one time.
This is the equipment we have:
1x Cisco Wireless Control Server (WCS) 6.0.181.0
4x Cisco 5508 Wireless LAN Controllers 6.0.196.0
60x Aironet 1142N Lightweight Access Points (LAP)
Is there a hard or recommended maximum number of clients per LAP? If so, where is this defined? From what I have read on these forums, Cisco apparently recommends about 25 clients per AP but I can not find any official documentation to support this.
When I go to WCS Home > General > Top APs by Client Count, the top AP reports 20 clients. However, if I click on the AP Name and go to the Current Associated Clients tab, it is only listing 8 clients - why is this?
Any help greatly appreciated.
Thanks in advance.

Answer 1

1.  What error messages do the clients get when they are having association/authentication issues?
There is no error message on the client, it simply fails to connect instantly. On an Android phone, you see it constantly trying to connect, failing and trying to reconnect. After many attempts (at least 10) it may connect successfully. This appears to me that it may be related to aggressive load balancing - here is what I think may be happening:
Client tries to connect to nearest LAP
Nearest LAP has to many clients on associated (as detemrined by the RRM Thresholds > Max Clients setting) and so refuses the connection (and sends a code 17 response)
Client attempts to connect to the next nearest LAP which is also too busy. It too rejects the connection.
Client attempts to connect to the third nearest LAP which is also too busy. It too rejects the connection.
And so on until it comes back to the first nearest LAP
This LAP, which still has too many associated clients, rejects the connection for a second time (its RRM Thresholds > Min Failed Clients is set to 3)
the clients tries the other APs and is also rejected as they are all set to reject connections for 3 times.
Client goes round all the APs for a third time. still getting rejected.
It then attempts a 4th connection to the nearest LAP and is allowed to associated as it is beyong the 3-attempt threshold for rejected connections.
This is my understanding of Aggressive Load balancing and would seem to match the observed behaviour of the clients but I would like to know if this is correct and how I can verify that in the logs?
This goes back to my first question, what is the max recommended number of clients for a 1142N LAP? Furthermore, is there any supporting documentation for this recommendation? Also, if I increase the RRM threshold Max clients to more than its default of 12, will the LAP accept more connections before rejecting them?
I would also like an answer to my second question in the original post.
2.  Are you using ACS/RADIUS/TACACS?
We are using 802.1X with ACS for our corporate users on the corporate SSID but the behaviour described above is seen on guest clients that just use web passthrough (i am not sure if it is also on corporate clients, I have not seen any reports to that effect yet).