WLC 2504 & LWAP AIR-AP1832I-E-K9 client throughput issues

[donal.carroll]

I recently migrated our Wifi system from old Aironet autonomous AP's to a centrally managed WLC system with 15 AP's broadcasting 2 WLAN's for guest and secure Wifi access. Both SSID's are using WPA-PSK authentication.

I've seen unusual problems with a small percentage of connecting clients. Clients authenticate fine and get an IP address, I can ping the default gateway and also any IP external address. There is no problem with DNS lookups either. However for some bizarre reason, clients browsers are incredibly slow. This is happening across different parts of the building (different APs) and not for all clients only a small number of clients, some using Windows (mainly Win10) and one using Ubuntu. When I tried debug the client on the WLC I don't see many messages other than connection and disconnection messages

(Cisco Controller) >debug client <mac-address>

What's even more unusual is that I booted the client laptops from a Ubuntu live USB and they have no problems with connecting. So that lead me to think, it was the Windows drivers. After updating the driver there is no change.

Also the problem itself is intermittent, some days it would be bad others it would be fine. The same clients have no connection issues on other wireless networks.

Every client connects on the 5Ghz 802.11ac band and can ping default gateway, external addresses and perform DNS lookups. 

There are no policies set on Wifi traffic. Other than the WLAN setups on the WLC 2504, everything else is set to defaults. Anyone ever seen this type of behaviour before?

[Leo Laohoo · Accepted solution]

 not for all clients only a small number of clients, some using Windows (mainly Win10) 

Are the clients running Windows 10 Creators Update, by any chance?

[donal.carroll]

2 of the clients I checked today are running the creators update. How would this effect our particular wireless network and no other network?

Also, I see similar symptoms on one Ubuntu machine. 

Is there something particular about this Windows update that might cause problems with a setting on my network that might be enabled and not another network?

[donal.carroll]

Ok, thanks for the pointer. So the problem seems to be surrounding MTU size on the clients and fragmented TCP packets. You can disable RSC on Win 10 by following the workarounds found here

https://answers.microsoft.com/en-us/windows/forum/windows_10-networking/wifi-issues-with-creators-update/4a20ba4f-33dc-4397-9823-e12dcb2607ba

I've tested this on Windows clients and the solution above works

However at the network level you might be able to solve this by enabling TCP MSS

http://www.cisco.com/c/en/us/td/docs/wireless/controller/7-4/configuration/guides/consolidated/b_cg74_CONSOLIDATED/b_cg74_CONSOLIDATED_chapter_01111110.html

Wireless -> Access Points -> Global Configuration -> Global TCP Adjust MSS enable

I'm going to test with the default 1363 but I'd say anything below 1500 should work. This should prevent the requirement to patch every client running Win 10.

Still interested as to why Ubuntu client has the same symptoms. Possibly similar MTU related issue. I should have broken out wireshark earlier ;-)

I will revert if this solves it for unpatched clients