Wireless clients sporadically getting Limited Connectivity, APs dropping packets

[John_Rowler]

We have a pair of V2110 wireless controllers and around 400 3825i access points. Controller firmware is 09.21.06.0002. AP firmware is 9.21.27.1387X.

About 4 months ago we noticed a problem where the APs would just drop traffic. A packet capture will show a ping coming into the radio from the client, going out the Eth interface on the AP, the reply coming back into the Eth interface, and never going out of the radio.

A particularly strange detail is that this only happens with certain source/destination traffic. For instance a client cant ping the gateway of 10.1.20.1/24. but cannot pint the server at 10.1.20.2. Since we can see the ping making it to the server and the ping reply making it back to the AP, we know it's not a firewall/ACL/routing issue on any other part of the network. The traffic disappears at the AP. Clients cannot communicate with anything on the Internet or any of our servers (DNS, DHCP, etc.) during this time. They can ping gateways in any subnet but that's it; everything else is failed.

More details:

-We've gone through several firmware updates with GTAC. Nothing has helped.
-The issue can affect any client on any AP in any location at any time.
-The issue is intermittent; it may happen every few minutes or once per month to a particular client in a particular area. Once failed it may last for seconds or days.
-While it's affecting one client, other clients on the AP may be working fine or may also fail.
-It can happen on either 2.4 or 5.7GHz.
-It can affect any type of client (phone, laptop, tablet) and any OS.
-Restarting the client or AP will sometimes fix the issue, sometimes not.
-It happens with our without using NAC, or any other type of authentication. It happens on completely open networks as well.
-It happens on both bridged at AP and bridged at controller topologies.
-It's not limited to a particular subnet or VLAN.
-It happens with or without Flexible Client Access enabled.
-It happens on either controller.
-GTAC hasn't made much progress on the issue in 4 months.

[Andre_Brits_Kan]

I stumbled upon this Gtac article
At this stage it would seem that the whole 9.21.x and 10.01.x has got problems....

[FES]

the utility of 802.1x and radio preference is that all is transparent to the user. At university context this is a adventadge. we cant create eduroam24 and eduroam5 ssid because is an ssid formely created to all the academic organizations.
thanks

[JP4]

What version do you have to be running for probe suppression to work correctly ? Earlier we were told to turn it off, now it sounds like it should be on ?

Are the problems everyone is seeing now related to disconnections from the network completely or still the limited connectivity issue that started the topic ?

I have been testing 9.21.09 on some AP's, but my client population is really low right now so its hard to tell of any issues.

[FES]

we have testing in some aps this morning, and in two hours no problems. We wil testing some days...

[Raffi]

Turning off arp proxy, using multicast to unicast checkbox, reducing size of l2 subnet or filtering out or containing to vlan some broadcast traffic should help. The issue is a client issue with apple products not waking up each dtim period so missing traffic.



http://www.sniffwifi.com/2016/05/go-to-sleep-go-to-sleep-go-to-sleep.html



Would be nice to know if there is a secret way to disable power mgmt. on os x.