Ghosts in our 6 New Ubiquity Pros - provision issues.

Bob Evans bob at FiberInternetCenter.com
Fri Jun 19 16:26:42 UTC 2015


> That's possible but I if they are re-provisioning on a regular schedule I
> kind of doubt it.  It would be easy to test though.  Plug an AP directly
> into your switch with a quality pre-manufactured patch cord and see how it
> acts. If it exhibits the same symptom it is probably not cabling.   Also,
> have you checked your interface counters for any packet errors?

Yes, no packet errors crcs or frags.

> Don't
> forget to look at your controller because if the controller became
> unreachable for any length of time that could easily cause your APs to
> re-provision as they reconnect with the controller.

This is did not know - thought the controller was just to provision and
monitor. After all why would a manufacturer make one point of failure for
a campus setup that uses thier own edgerouter for the dhcp etc. Doesnt
seem correct. But will will investigate it.

> I might set up a ping
> every second from the site of the access points to the controller and make
> sure the availability of the controller is 100%.

Yes that and what the ciscos report on the port link.

>  If you are on Cisco
> switches you should have log messages regarding PoE be granted on
> particular ports as well as up down messages on the interfaces.

Yep and we get them.

> Do you
> see the ports going up and down?  It is important to have NTP on the APs
> and switches so that you can correlate events in time (i.e. did the AP
> reboot causing the Ethernet link to drop or did the link drop causing the
> reboot?)

I am sure its the APs dropping - as non of the other devices VOIP phones
etc drop in the logs.


Thanks Steven
Bob
>
> Steven Naslund
> Chicago IL
>
>
>>Bob,  I've deployed tons of Ubiquiti gear, and have seen this problem
>> before. It always turns out to be poor quality cable installation. POE
>> does not tolerate low quality connectors, especially in outdoor
>> environments. There are >many aspects to a quality cabling job, so the
>> best thing you can do is seek out a qualified installer with outdoor POE
>> experience.
>>
>>The most common problem I see is people using crimp-on RJ45 connectors
>> directly on the ends of their cable runs. This is not how structured
>> cabling is designed to work, in particular because most crimp-on
>> connectors are intended for >stranded copper wire (such as that used in
>> very flexible patch cords, designed to run horizontally over only a few
>> dozens of feet), whereas the "riser" and "plenum" cable used for
>> long-distance runs has solid core wires. The tiny >teeth in standard
>> crimp connectors are designed to penetrate stranded wire, to make a solid
>> electrical contact. With solid core wire, they just bend to the side of
>> the copper core, making tenuous contact, which will conduct POE >current
>> poorly (resulting in the resets you see) and eventually fail altogether
>> as the improper connection corrodes over time.
>>
>>The correct installation process is to use "punch-down" RJ45 jacks at
>> each end of the cable run, and connect from those jacks to your equipment
>> (radio at one end, POE switch at the other). On the outdoor side, the
>> jack/plug junction >needs to be in a NEMA weatherproof enclosure, with
>> weathertight fittings. And, for human and equipment safety, you must use
>> shielded Cat5e/6 cable anytime you go outdoors, grounding only one end
>> (usually the radio end), and >protecting the cable with an inline
>> lightning protector between the RJ45 jack  and the radio.
>
>>If you haven't done that, then that's the first thing to fix.
>
>>BTW, avoid homemade patch cables whenever possible. Quality factory
>> cables are hydraulically pressed and the plug is hermetically fused for a
>> vastly superior connection compared to anything you can do with simple
>> hand crimpers. And >all outdoor cables must be UV-grade cabling with
>> weatherproof sheathing and water repellant inside (so-called "flooded"
>> cable).
>
>> -mel beckman
>
>





More information about the NANOG mailing list