AP3802 On 8.5.140.0 Intermittent Randomly

[fdharmawan]

Hi Guys,

I have Cisco AP3802 in production and recently upgraded my WLC to 8.5.140.0 because of some bugs. But, after the upgrade some of the APs sometimes became unresponsive to serve clients for several seconds. I'm not sure what happened but the AP was not down (uptime is not reset) and client still connected to the AP (but when pinging the gateway the result was RTO). I already checked the log both on WLC and AP, so far got nothing on that timeframe. I also checked the event viewer on the Windows 10 laptops, and also got none. I checked the interference from other AP (from WLC dashboard), it was minimal.

Any idea what else should I check? Because I'm kind of clueless here what's causing this.

Thanks.

[Ric Beeching]

Sounds like a bug my friend. Do you have TAC cover? If so please log a request so it gets captured by them. Before doing this, confirm if it only affects specific clients or all types e.g. windows intel chipset vs iphone.

Cheers,
Ric

[fdharmawan]

Hi Ric,

Yes I do have TAC support, but I'm not sure what to give to TAC since the logs on WLC and AP were clean. Any advise where should I start?

[Ric Beeching]

Tricky as you aren't seeing much evidence. Can you replicate with a client or you just have to wait for it to happen at some point? If you can replicate, try a packet capture on the device using wireshark and give to TAC along with a 'debug client <mac>' from the WLC for that client.

Also worth letting TAC know if you're running FlexConnect or not as i've seen a few 3802 specific bugs indicating some silent drops.

Ric

[fdharmawan]

Hi Ric,

Unfortunately I have to wait until it happens again. Regarding packet capture, should I use the additional USB packet sniffer? Or can it accomplished by sending the logs on the AP to a syslog server?

[patoberli]

I would send the logs to a syslog server.
Do your problematic clients have a new wi-fi adapter that is MU-MIMO capable? And if yes, do you have it enabled on the WLC?
There is a bug in 8.5.140.0 that some Intel MU-MIMO clients might lose the ability to transmit data if both are enabled. In that case, disable MU-MIMO on the WLC for testing. Will be fixed in the next 8.5 release.

[fdharmawan]

Hi Patoberli,

Most of our wireless NICs are MU-MIMO capable. And how do I disable MU-MIMO on WLC? Tried to google it and search it on WLC, but so far no luck.

Thank you.

[patoberli]

You actually configure this under the SSID - Advanced.

[fdharmawan]

Hi patoberli,

Ah finally I find it. Thank you very much. Will disable it and monitor for the next 7 days.

Can I get the link to the bug page that you said before? Just for my reference and my justification to the management.

[patoberli]

Had to search a little, but finally found it:
https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvn66715

Check if the text under "Symptom" applies to you.
Please note, I don't have any issues with 2802i APs (which normally run the same software and have the same hardware as the 3802i) with 8.5.140.0. At least no user reported an issue so far and running that code since several weeks.

[fdharmawan]

Hi patoberli,

The symptoms are not the exam same. I did not find the exclamation mark on the WiFi icon and I do not have to disconnect and reconnect (reassociate) since the traffic will be normal after waiting for several seconds. Maybe the root cause is the same? Kind of a similar thing.

I do have 2802 on deployment, but as WIPS. And the WIPS is not functioning as expected on 2.4GHz band.

[patoberli]

I think it's worth a try.

I can't comment on WISPs, I never used it, because I don't have an MSE.

[fdharmawan]

Do you have any concern for disabling 11ac MU-MIMO? All of the laptops on our environment are 11ac capable. But there are mobile devices that are not. The comparison let's say 1:1.

[patoberli]

MU-MIMO actually requires 802.11ac Wave2 clients that are MU-MIMO capable. This list is usually very short. I would even bet that you don't see any apparent difference with it disabled.

[Reginald Pugh]

Is this issue for APs in Local Mode Centralized traffic to the WLC or is this a FlexConnect and local switching?

Am seeing some issues with multiple sites where some clients can't connect to an AP even when standing underneath it while other clients can (bluelight). We are going to get a packet (frame) capture to see if we can get the client details. But are interested if we are hitting a variant of this since we are on code 8.5.130.0 since beginning of March 2019. Now that we have more APs and sites, th issue is becoming more common.