WS6 Mini - port state chnages every 30 secs or so

[rodrayner]

Hi @ sirhc

The AP = Ubiquiti Outdoor AP Plus. Latest controller and firmware on the AP

Nano's are all NanoStation M5's - all now running V6 firmware.

All Cat5 cables are factory produced, we do not make any cables ourselves.

The switch is definitely auto negotiate, i have to presume the AP is as i do not know of any way to alter this and looking at the UBNT forums everybody is pretty much saying there is no way to change it either ( although i expect its possible via SSH which i haven't done ). Just to note, when installing the AP's are taken from the unopened box, mounted to the column and powered up via POE. when the network is up and running they are adopted into our cloud controller and the site config takes over.

I have ran a cable test which has proved to be interesting, the results are below..

And when i first saw then i was concerned that i had bad cables... but we are getting the same port state change on every switch.. so i ran the cable test on every switch... and got the same result on every switch... I also ran it on some of the ports running the Nano's... got the same result every time i ran the test.

The pair length has to be incorrect as i installed all this kit and i know for a fact that the fly leads connecting the kit to the switch are 1mtr, in fact some are only 0.5. It also suggest that only two of the pairs within the same cable are different lengths... With this in mind I'm not sure the other results of bad cable and short status are correct either

Finally... regarding cables, taking into account every fly lead we use if factory produced, as we are getting the same error on every switch... on a single port in the main ( a little on others ) ... i can't believe its a cable problem

So moving on.. found something odd... looking at the logs on all switches this morning... all the switches that are slightly away from the main town area stopped reporting an issue around 8:30 pm ( GMT )... the two that are right in the town center stopped reporting issues around 9:30pm

Nothing has changed other that the one switch has had a firmware upgrade but that was later at around 10:30 pm from memory, so i am presume the logs stopped due to inactivity of client connections which is very likely as this location is not very busy, especially at night.

I will keep my eye on it throughout the day and let you have an update after lunch when i know the town would have been much busier.

In the mean time any info re the cable test would be appreciated. I may grab a new switch from stock, couple of Nanos and AP and set it up in the office to see what cable test i get from that.

Thanks for your help so far.

[rodrayner]

Hi, Just a quick note.

I have just checked 4 other sites we support and every switch i checked has the same symptoms. Some worse than others but in essence they all have log errors full of port state changes.

Do you have a support system that you would prefer to move this issue over to ?

Thanks.

[sirhc]

The cable diagnostics is NORMAL for UBNT airmax

Read this post: viewtopic.php?f=6&t=1178

Question: Why do I see short on pairs 3 and 4 on a of devices especially UBNT airMAX 10/100 gear.

Answer: Because every manufacturer does not follow the proper design specs and use a 4 channel Ethernet Transformer designed for POE with a center tab for each pair. Since pair 3 and 4 is not needed to carry data with 10/100 only links they use them for power only which allows them to use a cheaper 2 channel transformer without a center tap for pairs 1 and 2. They then connect the 2 wires from each pair to the board skipping the need for a transformer with a center tap so the 2 wires from pair 3 are connected directly to the board shorted together for positive and the 2 wires from pair 4 are shorted together directly on the board for negative.

All Cat5 cables are factory produced, we do not make any cables ourselves.

You should be using ESD Cable and Ends

[sirhc]

I have just checked 4 other sites we support and every switch i checked has the same symptoms. Some worse than others but in essence they all have log errors full of port state changes.

Do you have a support system that you would prefer to move this issue over to ?

I am going to go out on a limb here and say there might be some interference on these light poles interfering with Ethernet communication?

There is nothing wrong with the switches as there are THOUSANDS of these out there in service working fine.

If you had 1 acting weird it could be a bad switch but since you have all acting the same it has to be something with the sites or the way you are using them that is the issue, not sure what this is.

Whatever it is it is messing with Ethernet communications as there is a momentary blip where the Ethernet link is dropping but the devices nor the switch are rebooting.

So when you say pre-made jumpers are they shielded?
Do they have ESD ends and drain wire?

[sirhc]

Yes there is a setting on all UBNT radios where you can set the speed duplex in the UI.
I would login to all the devices and make sure they are all set to AUTO.

I would also take this time to upgrade all the firmware on ALL the devices.

Do this is a great big long daisy chain setup where you start at one light pole and feed the next then the next then the next correct?

How many hops deep is this chain?

[rodrayner]

@sirhc,

I appreciate the feedback and i know there are thousands of the switches out there... i personally have installed 300-400 hundred... and was one of the first people in the UK to get them when you released them to market around 18 months ago. In fact no one had them in the UK so MS Dist brought them in specially for me from you.

I get what your saying re potential interference ( although not convinced ) but as i mentioned on my last post i have checked EVERY location i have these installed and EVERY site has the same symptoms. So put that in perspective.. there are thousands out there and i can show you around 300 or so that ALL do the exact same thing!

I don't mean to sound moody but it sound likes the config it good, everything used seems to be correct but as we can't put our finger on the reason its happening.. but.. it still can't be the switch...

Re the issue with AirMax kit... does that also apply to Unifi... we are discussing in the main AP's here and not the Nano's

Having said all that... i have re-checked all the logs and unless they are not refreshing.. there are no reports listed.. well not many anyway and i'm totally cool with just a few.

This is the log so far for the same switch mentioned previously in this discussion

Thanks

[Julian]

Pretty sure we're fine here; if we come to some discovery that helps, we'd like it to be easily searchable for future clients.

I'm leaning towards the ethernet cables coupling AC interference as well, since most factory cables aren't well shielded, nor do they have dedicated drain conductors, and they're probably running up the inside of the light pole right next to the AC lines for the lights, unless my guess is missed.

The nanostations have a more robust PHY (networking shorthand for physical interface) than the UAPs do; this could be why you're only seeing it on that particular port.

re: not seeing it at this current moment, are the street lights on?

[sirhc]

Re the issue with AirMax kit... does that also apply to Unifi... we are discussing in the main AP's here and not the Nano's

All UBNT 10/100 devices will show SHORT on pair 3 and 4, it is the way they made them.

It was not industry standard and should not occur which is why we say fail.

As far as the cable diagnostics showing incorrect length on the 1M or less cables the cable diagnostics is not that accurate with +/- 1M accuracy and when the device is not powered or the the pairs are wired directly into the board as they are with UBNT 10/100 devices on pair 3 and 4 this will mess with pair lengths on pairs 3 and 4.

[sirhc]

I think you have an excellent idea.

Setup a LAB in your office on a bench and see if you see this Ethernet drop. i am sure you will not.

Also investigate the radios to make sure someone did not hard code them to 100M which would cause this issue.

All ubnt airmax radios allow you to set speed/duplex in the web UI.
In the UniFi it will be set in the controller.

If the ports are hard coded to 100M this is your problem and is an easy fix. < If not interference on the cable this is your BEST BET

[rodrayner]

Guys, thanks for all the feedback and the speed in which the feedback is given... much appreciated, Julian, welcome to the discussion

Re the power interference.. i totally get that and in all honesty never occurred to me until you just said... i will have to find shielded cables pre made for the inside run as we install loads of these and i HATE... making up fly leads..

The lights are not on at the moment... but all the logs seemed to have stopped late last night.. when they were on...well i presume so as some street lights in the UK do go off around midnight to save on money... things are tight with Brexit you know..

I think best plan moving forward is to build the test rig. i won't be able to replicate the traffic but it will hopefully provide some data.

Give me a few days and i'll update the notes after some further testing.

Re the auto negotiate... i can not for the life of me see anywhere in the controller or individual AP config ( GUI ) where you can change it.. i expect it is possible via an SSH session but i woudl be happy to bet my house that no one has done that here... will check though.. lets hope i'm not made homeless...

Cheers guys...