Unifi Wireless constantly dropping connection

[TMBS]

Hi

 

We have 29 Unifi AC Pros around school that have been randomly dropping connection over the past couple of weeks giving the following errors:

 

Missed heartbeat

High TCP latency for Client XXX

High CPU usage for Access Point XXX

Client XXX is having trouble resolving a domain name to an IP address (DNS timeout).

Client XXX is having trouble resolving a domain name to an IP address (DNS timeout).

 

Usually a reboot resolves the issue, or they just come back online but over the past couple of days it's gradually getting worse. It's not the same ACs, it just seems to randomly happen across our site. The Chromebooks say they're connected but also gives a network not available and users can't login, or browse the Internet.

 

At the moment we have 8 Chromebooks that can connect and 34 that won't connect, all in the same classroom!

 

We have a flat network, no VLANS, DHCP and DNS on our domain server running Windows Server 2016. Some APs are EOL and we're trying to find a good time to replace...

 

Any ideas of what we can do?

 

Thanks

[caffrey]

What gateway do you use ? A USG ?

 

Other thing to try is SSH into an Access point, and using TOP to see what's using the most resources and go from there

[Driftingashore]

Not going to be much help, but you'll probably be better off asking in the Ubiquiti community forums. We've recently been having reliability issues with DHCP leases not being granted / renewed for clients on Unifi APs until they're restarted (again.. this is probably the third time). If we weren't so far into Ubiquiti hardware I'd ditch them in a second given I can't depend on them to release firmware that doesn't break everything.

[caffrey]

I'm due a refresh but I don't know of an alternate - a topic for another thread I think.

[TwistedHelixis]

There were a few posts recently on Edugeek with similar issues, so I would read through those.

 

I had a similar issue which seems to have fixed itself after updating Chrome OS, the AP firmware.

 

Something else that may have helped but I have no way of knowing, I created a new SID with 5ghz disabled, all the Chromebooks that were not working seemed to connect fine, I then switched those Chromebooks back over to our main SID and they all worked OK.

[caffrey]

Mine was DNSMasq and avahi-gateway - I suspect a BYOD doing something it shouldn't but I couldn't trace it, turning off an option and moving avahi to another server fixed all my issues overnight

[ITGURU]

Have you upgraded to the latest firmware? there was some DHCP issues on a buggy firmware - upgrade fixed it for us.

[mrwoberts]

Please could you detail the firmware version(s) you are using, plus the controller version?

 

Also, could you detail (without sensitive info) the settings that you have enabled/set on your Wireless Network (block LAN-to-WAN, fast roaming, Rate limiting...)

Are you using DFS channels on the 5 GHz band ? (channels 52 onwards)

Are you using Band-Steering or Cell-Size Tuning at the AP level ? - I'm not 100% sure that's available on AC Pros

Are you using the Wireless AI feature ?

Edited May 20, 2021 by mrwoberts

[TMBS]

Hi everyone

 

Thanks for your replies! Dad a manic day yesterday. Can't pinpoint to a specific AP, device or wifi network, so really hard to work out what the issue is. For now I've updated the Controller, Cloud Key and AC firmware so they're all on the latest versions.

 

We don't have the Unifi Gateway.

No we're not using DFS channels, Band Steering or Cell Size Tuning

We weren't using the AI feature but I've just turned that on (and reading up on what it does)

 

I really need to get some experts in to help us set it up correctly and I presume it would be better on a VLAN? SLT don't want to spend the money though...

[TwistedHelixis]

Is it just an issue with your Chromebooks or do other devices have the same issue?

[TMBS]

Is it just an issue with your Chromebooks or do other devices have the same issue?

 

All devices.

[chazzy2501]

VLANs control broadcast traffic, so don't sweat them so much unless your network goes down in a broadcast storm (then worry) I feel your pain my new HP Probook 440 G8 (11th gen intel with AC wifi) just are having poor performance compared to our older laptops that only have wifi n.

 

I have the HD NANOs with a Feb firmware on them. I actually get the laptops either performing much slower than they should roaming poorly and two of them declared that the wifi key was invalid until I powered the laptop down!

 

Unifi are the devil I can afford.

[TwistedHelixis]

Have you checked you have enough IP addresses left in DHCP?

 

Are there any duplicate IP's in DNS?

 

Although it does sound like an AP issue I would rule out the above

[Oaktech]

Had something very similar...

 

One of the guys from my MAT fixed it with thefollowing settings:

 

DTIM set to 3

Ensure LAN details are correct - it looks like it only applies to networks with a Unifi USG, but it can cause DHCP issues.

Allow BSS transition with WNM

PMF Disabled

GTK Re-key 3600 seconds.

 

Latest controller and firmware versions.

[TMBS]

I think my technician has found out the issue. We have a Cloud Key Gen 1 controller and they only support 23 devices...

 

Need to decided whether to install the controller software on our server, or buy a Gen 2!

 

Thanks everyone

[Michael]

There isn't a hard limit, only a suggested recommendation.

 

The Cloud Key Gen 2's are more powerful and better made than Gen 1.

[localzuk]

The issue is less the number of devices controlled, and more the volume of data being sent back - if you have 23 devices spread out and only have 50 clients connected to them, you'll be fine.

 

We found a cloud key 1 to be rather weak - it struggled once we hit a handful of Unifi switches. Ended up upgrading to a UAS-XG, and now over to a virtual machines running the controller instead.

Edited June 8, 2021 by localzuk

[Oaktech]

The issue is less the number of devices controller, and more the volume of data being sent back - if you have 23 devices spread out and only have 50 clients connected to them, you'll be fine.

 

We found a cloud key 1 to be rather weak - it struggled once we hit a handful of Unifi switches. Ended up upgrading to a UAS-XG, and now over to a virtual machines running the controller instead.

 

Interestingly we had guest portal issues with running the controller directly on a Windows VM which was fixed by moving it to a docker implementation on windows (along with our Xibo instance) with linux containers via the linuxserver.io image.

[jthompson]

Cloud Key Gen 1 is only a 32bit device, so once the UniFi database gets towards 2GB you get all sorts of problems with APs not being able to submit statistics to the controller. Later Cloud Key hardware is 64bit so the limitations on db storage aren't such an issue.