Delayed roaming

[Pako1]

I have inherited a 5 year old WLAN deployment where clients have never successfully roamed. There is a 10 - 20 second delay when moving between AP. During this break the wifi icon disappears from devices and connectivity is lost. We have a mix of iOS, Android, SurfacePro and the odd laptop on the network and all encounter the same issue. I do not see any disassociation events in the debug when this happens.

Where do I troubleshoot from here to track down the root cause? have attached a debug log

5508 (8.2.151.0) controller and mix of 3502 & 3602 AP. Dual 2.4/5Ghz

[michaelbs]

Hi,

I have a similar question. We are running two WLC 5508 controllers with software release 8.3 in a HA configuration and Cisco 2802E 802.11ac Wave 2 access points.

All access points placed in the same building are attached to the same controller so that Intra-Controller-Roaming does not occur when moving in the building.

However, we sometimes observe that clients (mostly Windows notebooks with Intel 8265 wireless NIC) stay too long connected to the same AP, whereas they should already have moved to another access point that is closer.

We never have configured anything special regarding to roaming. However, is there something we can do in order to increase the roaming time to make sure clients roam to a closer AP more quickly?

Thanks
Michael

[patoberli]

What is the roaming sensitivity configured in the wlan driver?
And is the driver more or less up to date?
Don't forget, the client decides when it wants to roam.

[michaelbs]

After Reading the following blog post about sticky MacOS wireless clients:

 https://framebyframewifi.net/2017/08/20/macos-wi-fi-roaming/

I have enabled Optimized Roaming in our environment:

  https://www.cisco.com/c/en/us/td/docs/wireless/controller/8-1/configuration-guide/b_cg81/b_cg81_chapter_010001110.pdf

I'm curious about whether this will help! Will let you know in a few days.

[patoberli]

From the Optimized Roaming manual: "This feature disassociates clients based on the RSSI of the client data packets and data rate."
Because of this, latency sensitive applications will have outages, also most older drivers don't understand what's happening and will cause an up to 1 minute outage. Also it can sometimes take a few minutes until a client can connect, because it tries and retries to connect to an "not optimal" AP, but always gets denied.
Use it with caution!

[michaelbs]

Thanks. We will test and observe this. Fortunately, all our laptops are quite new and therefore have new chips and wireless drivers install. It it causes issues, we will disable it.

[Rasika Nayanajith]

First I would go to 8.2.170.0 code & see if behavior remain as it is.

HTH

Rasika

*** Pls rate all useful responses ***

[Pako1]

The WLC/Aps  has been through many version over the years and none have made a difference.Is there a reason for you thinking upgrading to that version will resolve this? 

So if the clients are sticking to previous AP, how can I validate that is indeed what is happening? On the iphones it appears disconnected but debug shows a reassociation not disconnection/association. Is a 20-30 second (sometimes minute) delay normal in these cases?

How can I tune the system to make the clients more likely to decide to roam faster instead of hanging on to an AP that it cant communicate with. Can I adjust rssi so signal strength between AP is lower? 

[Rasika Nayanajith]

Is there a reason for you thinking upgrading to that version will resolve this? 

No point spending lot of time in troubleshooting, if current code you running have known issues. As of today 8.2.170.0 is the recommended code in 8.2 train. So let's do that first & then spend time on troubleshooting if issues is there.

Is a 20-30 second (sometimes minute) delay normal in these cases?

It is not normal taking that long to roam.

Pls give more information about how your WLAN is configured (WPA2-PSK or WPA2-Enterprise, what encryption methods, etc), is any fast roaming methods configured (like FT)

HTH

Rasika

[patoberli]

If you have a dense enough coverage, you can lower the signal strength of the APs and also increase the minimum speed (disable rates 1, 2, ... and leave 18 as the minimum enabled).

[Pako1]

Lower data rates are blocked. Density is pretty high in some sites so could be backed off. sh run attached

[patoberli]

Not fully. On 802.11a is 12M still enabled, although that is fairly ok and I wouldn't change it at the moment.

If I read the config correct, you "only" have 20 MHz channels on 5 GHz, right?
That actually means your clients will not see an improved throughput on 5 GHz compared to 2.4 GHz. If your channel plan allows, enable also 40 MHz channels, that can double the throughput and will "look more interesting" for the clients to attach to, even at a slightly lower signal strength.
Check on the Home Tab of the WLC how the channel load on 5 GHz is during a normal day. If all the APs are < 10%, then you can probably savely enable 40 MHz. But don't forget to check it again after you've made the change.

[michaelbs]

I have enabled and tested Optimized Roaming and it seems that it is indeed makes roaming faster. I just tested with a new Windows 10 laptop with Intel 8265 and wandered through the building. Each time I crossed another AP the client quickly associated with the AP.

Should be enable CleanAir in order to make things more smooth? We currently have CleanAir completely disabled, however, all 2802E access points are capable of this.

[patoberli]

If you use RRM, DFS, and so on, yes it would benefit from using CleanAir.

[michaelbs]

Thanks.