[27209] in resnet
Re: Ethernet ports "burning out"
daemon@ATHENA.MIT.EDU (Osborne, Bruce W)
Thu Jan 26 08:00:15 2012
Content-Language: en-US
Content-Type: text/plain; charset="iso-8859-1"
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Message-ID: <7F8CAE21F9C1C94A90F11320EF3974CE25B11C60@LUEMSMAIL01.University.liberty.edu>
Date: Thu, 26 Jan 2012 13:00:03 +0000
Reply-To: Resnet Forum <RESNET-L@listserv.nd.edu>
From: "Osborne, Bruce W" <bosborne@liberty.edu>
To: RESNET-L@listserv.nd.edu
It is obvious that PoE would not work in the split pair situation. If your switch is PoE capable, is it turned off?
Bruce Osborne
Network Engineer
IT Network Services
(434) 592-4229
LIBERTY UNIVERSITY
40 Years of Training Champions for Christ: 1971-2011
-----Original Message-----
From: Brad Coburn [mailto:coburn@tcnj.edu]
Sent: Wednesday, January 25, 2012 9:08 PM
Subject: Ethernet ports "burning out"
Hi all.
I have a curious situation that perhaps the community might be able to offer some insight on. I apologize for the length of my post; I hope the detail will serve to spark some insight.
Some background:
Two, 10-story residence halls. One Ethernet port per student (so 2-person rooms have two Ethernet ports). Three equipment rooms per hall, serving approximately 1/3 of each building. 15, 48-port 10/100 switches total in each building.
The situation:
Suddenly this Fall I have a recurring issue with a single student room.
The initial report is that they cannot establish connection with their laptops. One is a PC, one is a MAC; neither looks terribly old (I'm on the network side, not the desktop side).
First time, troubleshooting turns up "bad" Ethernet ports on that switch for these two connections. Not an uncommon occurrence across residence halls. If it's important, the two patch cords are adjacent (jacks were patched in, in-order).
Moved patch cords to two spare ports (happen to be adjacent) on that switch; service restored (confirmed with my tester and customer machines).
Two weeks later (second time), call comes back. No link. Troubleshooting shows two "bad" Ethernet ports again. There are only three of us who work on switches in the field, and nothing's been touched (so the patch cables are where they were). Moved both patch cords to another switch, adjacent ports, same model.
Some days later (third time), call comes back. It's "bad" ports again.
So I and my tech go out there. Standard tests come back fine.
Re-terminated the jack "just in case". I brought my cable pair tester to try advanced functions like TDR, as well as resistance and voltage measurements pair-to-pair and pair-to-ground, with computers connected and without computers connected. Nothing comes up.
By now I'm out of sufficient spare ports, so I place these two connections on a little desktop Netgear switch and patch that into the stack. This is just before Thanksgiving break. Student connections again confirmed working with tester and customer computers.
Shortly thereafter we replace the two switches with "bad" ports (there had been other port failures over the years) with two new switches from spare stock. One switch is the same model. One switch is the newest model of the 10/100 class - a model I'm placing in other buildings that won't support or can't afford GigE.
The Netgear setup worked the whole time, but it's a non-standard setup and will only come back to haunt us later. So we pulled the Netgear out and placed the room patch cords back in their usual spot, which is now also on the new replacement switch.
Everything worked fine from approximately post-Thanksgiving Break until last Thursday, January 19. Most students returned from Winter Break on Tuesday January 24.
Call is back (fourth time). Ports are "bad" again, on this brand-new switch. This model supports extended diagnostics, which sometimes report bad ports. Nothing comes up this time (this is not atypical).
Tech and I followed the cable in the ceiling all the way to the IDF (only one floor down). This room is roughly in the middle of the cable runs, and the room next door shares a common route from the hallway.
Inspection turned up nothing in the way of obvious damage. Moved the customer patch cords to two spare ports on the new switch.
So tomorrow I have Facilities scheduled to go up and run new cable for this room. The parents are irate at this point.
I'm more than sure that this is a customer equipment issue, but I can't figure out what could be happening. We've not seen anything like this before. A single-port occurrence would be more likely rather than a simultaneous 2-port event.
It may be important to know that when the campus was originally cabled someone took advantage of the "economy" offered by 10Mbps Ethernet's operation over 2-pair. So we split the cable at the faceplate and then again back in the IDF. That's how it was done back then, and we're still operating successfully at 100Mbps now (but we stopped the practice for retro-fits some time ago).
Given our "split" architecture in this case, I suppose there's a remote possibility that a damaging event on one half of the cable could transfer to the entire cable. It's also just as likely that if customer equipment damages the first connection, the customer tries the second connection immediately and blows that too.
I certainly understand what this looks like for the customer, but I don't know what else to say. Of course the customers aren't doing anything out of the ordinary, and neither is anyone else they might invite into the room.
I have not tried:
- Swapping horizontal cables with the neighbor. If the problem follows, then it's ours. It's too late now though; Facilities will replace this cable and we will pull the old one down to more thoroughly inspect for physical damage.
- Separating the two patch cords for the room between switches. It just hasn't happened.
- Installing a surge suppressor on the line (I have pricing).
I really would like to know what I can do to damage an Ethernet port. By the way, my tester "sees" the bad ports as a switch connected, but powered off. Which should mean that it sees the transformer at the front of the interface. In no case has the tester reported any "error"
condition. Oh, and it's a Fluke LinkRunnerPRO.
Any thoughts?
Thanks.
-Brad Coburn
Associate Director, Communications Technologies Information Technology The College of New Jersey
15 switches
48 ports each
___________________________________________________
You are subscribed to the ResNet-L mailing list.
To subscribe, unsubscribe or search the archives, go to http://LISTSERV.ND.EDU/archives/resnet-l.html
___________________________________________________
___________________________________________________
You are subscribed to the ResNet-L mailing list.
To subscribe, unsubscribe or search the archives,
go to http://LISTSERV.ND.EDU/archives/resnet-l.html
___________________________________________________
