Unifi Access Points disconnecting

[gh5000]

Hi, we replaced our old Meru wifi over the summer with Unifi and have been having persistent problems with them disconnecting from the controller and being unpingable. They go disconnected seemingly at random and at different points of the day. Clients then can't connect to the WiFi AP but there is a SSID being broadcast. The situation is resolved by a PoE reset of the AP until it happens again.

 

We have 64 access points consisting of 10 AP-HD, 12 AP-AC-Lite, 2 AP-AC-Pro, and 40 AP-nanoHD all on version 3.9.54.9373. Looking at my tally of which access point has been going off to see if there's a pattern I think it is always the nanoHDs, but there is no other pattern of which edge switch etc.

 

Things we have done so far from the Ubiquiti forums:

1) Disable uplink connectivity monitor

2) Where we cheated a little bit over Summer and didn't swap the Meru PoE injectors for the ubiquiti ones to save time - this has now been done.

3) Any access points being supplied by a PoE switch have now had the Ubiquiti injector installed.

4) Fluke tested cabling for any faults.

 

The access points are untagged on their own vlan as I am short of internet accessible IP addresses. The controller running on Ubuntu has a NIC on the same vlan as the APs and the inform address is set to that IP address. The controller has a second NIC on an internet accessible vLan so that we can access the controller remotely.

 

Things I haven't tried yet is a factory reset of the access point and a readoption but does anyone else have any other ideas?

 

Thanks

Edited December 14, 2018 by gh5000

[gh5000]

Writing things down has helped. Once I realised it was always the nanoHD I had a look at the firmware for issues, it is on the latest but there are two more releases that haven't come down to my controller yet. The release notes for the next version have a lot of stuff related to the nanoHD so will try an upgrade or downgrade and see if the problems go away.

 

[nanoHD/IW-HD] Add client `Framed-IP-Address` to RADIUS accounting data.

[nanoHD/IW-HD] Improve 802.11k scanning.

[nanoHD/IW-HD] Improve Ethernet stability.

[nanoHD/IW-HD] Improve Fast Roaming support.

[nanoHD/IW-HD] Fix a bug which may make bandsteering fail to initialize.

[nanoHD/IW-HD] Fix instability caused by enabling WLAN scheduling.

[nanoHD/IW-HD] Fix throughput issue for clients with certain Intel wireless cards.

[nanoHD/IW-HD] Fix wireless adoption issue.

[nanoHD/IW-HD] Fix a memory leak.

[nanoHD/IW-HD] Fix multiple downlink support.

[nanoHD/IW-HD] Fix wrong passphrase issue that some client STAs encountered.

[nanoHD/IW-HD] Fix mDNS packets being dropped by IGMP snooping.

[nanoHD/IW-HD] Fix bandsteering support when WLAN scheduling enabled.

[nanoHD/IW-HD] Fix a bug causing instability when netconsole was enabled.

[nanoHD/IW-HD] Fix a bug which may cause some SSID(s) not to broadcast.

[nanoHD/IW-HD] Fix a bug in bandsteering.

[nanoHD/IW-HD] Fix an issue with wireless adoption when in factory default state.

[nanoHD/IW-HD] Fix wireless uplink VLAN broadcast/multicast packets format error issue.

[nanoHD/IW-HD] Fix SSDP dropped by IGMP snooping issue.

[nanoHD/IW-HD] Fix a bug causing a drop in transmit performance.

[nanoHD/IW-HD] Fix RF scanning support.

[nanoHD/IW-HD] Fix support for 8 SSIDs per radio.

[nanoHD/IW-HD] Fix a bug which may cause an AP to stop forwarding packets when configured to use wireless uplink.

[nanoHD/IW-HD] Fix a bug when using VLANs in a wireless uplink topology.

[nanoHD/IW-HD] Stability improvements.

[JOrdan01070]

Hi @ghanel I had the same issue for the nanohds, I upgraded the controller to 5.9.32 and upgraded the AP firmware to 4.0.10.9653 as that has a lot of fixes for the nanos. Do you have advanced features turned on. So far I have not had any issues for 2 weeks.

 

Jordan

[gh5000]

No, turned off advanced features a while ago as a potential fix. Upgrading the firmware now.

[evasion]

I have exact same issue, Unify nano HD access points "disconnect" randomly for no good reason. They are not responding to ping, SSID is broadcast but client cannot connect.

Controller upgraded to latest firmware, disabled automatic firmware updates for APs just in case they lock up if firmware is not good.

They can be reset by reconnecting to poe switch but that is not the solution.

Advanced features on controller disabled as radius authentication was failing randomly.

Any updates on "disconnecting" APs at your end?

 

thanks

[TwistedHelixis]

If you are using POI from a switch I remember reading something ( a while back) about the Unifi AP needing different voltages to the voltage supplied by most switches.

[Primus]

If you are using POI from a switch I remember reading something ( a while back) about the Unifi AP needing different voltages to the voltage supplied by most switches.

 

The Nano HDs use standard PoE

[evasion]

POE supply is not the issue here, all access points get 48V standard DC supply. Both cloud controller and APs are on latest firmware.

[StevieM]

I'm having the same problem! APs are now disconnecting at a rate of knots (currently 6 out of 22), and both switches disconnected yesterday afternoon. I've no idea how to restart them; networks* are not my strong point.

 

*Increasingly, IT in general is becoming not my strong point!

[caffrey]

Unifi AP's are weird, I have some with an uptime of 200 days+ yet I get some that go up and down like a yoyo

 

If they don't come back I usually restart by power cycling the switch port

[supportman]

Have you tried replacing the actual hardware transmitters? Nano's are really cheap. Edited September 19, 2019 by supportman

[evasion]

I have this issue raised with Ubiquity support via chat.

I only see them asking for logs, no resolutions at all. What's worse is that once "disconnected" AP is rebooted, logs from AP collected after that only show records since reboot, not when the fault happened. I get warning email generated by controller but that has no mention of why the AP became unresponsive.

[jslate1980]

Any you checked your switch logs? Are you using rapid stp? Had the same issue when using rapid spanning tree protocol. I think it was down to the wireless mesh feature causing a loop. Which caused the port to be locked out.

[evasion]

No issue with stp but I have been told to switch off "band steering" in advanced features menu. This looks a bit confusing setup. Advanced features are in controller menu, then when turned on, these features are available on each AP. So, turning off advanced features on the controller, does not turn off these features on AP-s, at least that is what my understanding is. Now, I have turned off "band steering" from balanced to Off on each AP, rebooted them, and then turned off advanced features on the controller. Fingers crossed. All AP-s on line.

[evasion]

Unfortunately this has not helped. Next day, 5 or 6 APs disconnected. Approx 10 (out of 51 total) have short uptime suggesting spontaneous reboots.

Ubiquity support on the case but not resolution yet.

[supportman]

Did you get this sorted in the end? I'm on the verge of pulling the trigger going completely unify for switches and wireless but this has me worried.

 

I don't think an AP on our Aurba system has ever gone down in 10 years other than for power reasons.

[Primus]

We've gone Unifi for WiFi and edge switching and have zero problems just to reassure you!

[supportman]

We've gone Unifi for WiFi and edge switching and have zero problems just to reassure you!

 

Great news, what models are you using?

[Primus]

UAP-AC-Pro and UAP-AC-HD - in total we have 95 APs.

With 48 port 500w POE switches and 48 port non-POE switches. - in total we have 38 switches.

[FragglePete]

I've got a small set of UniFi switches running our CCTV system, and I did have one 8 Port PoE switch do random disconnects continuously after working fine for sometime. I eventually did a full factory reset on it, re-adopted it and since then it's been rock solid.

 

I'm too am looking at their WiFi stuff - have a couple of the APs running as a test and impressed with it so far - have got dynamically assigned VLANs working on the cloud key with a Microsoft NPS Server and seems to work well.

 

Pete

[StevieM]

Power cycling* the switches brought everything back online for us last week. No problems since then.

 

*For power cycling, read pulling the plug!

[PotNoodleTech]

UAP-AC-Pro and UAP-AC-HD - in total we have 95 APs.

With 48 port 500w POE switches and 48 port non-POE switches. - in total we have 38 switches.

 

How old are the switches?

[mavhc]

Always seemed a bit dodgy that they had their own PoE thing and weren't following the standard to start with.

[Primus]

The POE switches are 1 year old, the none- POE switches are only a few months old.

@mavhc - they moved across to using standard POE quite some time ago now.

[evasion]

Update: Rolled back firmware to 4.0.42.10433 and have not seen any disconnections for almost a week. Goes to show latest firmware has a bug!