"test cable-diagnostics tdr" odd results

joebrug · ‎02-23-2017

Catalyst 2960-X switches running Version 15.2(3r)E1

Trying to track down why some devices are connected at 100mb instead of 1gb, like they should. So started using "test cable-diagnostics tdr interface gi1/0/1"

When device is plugged in, showing up/up at 100mb, I see the following result:

Interface Speed Local pair Pair length Remote pair Pair status
--------- ----- ---------- ------------------ ----------- --------------------
Gi1/0/1 100M Pair A N/A N/A Not Completed
 Pair B N/A N/A Not Completed
 Pair C N/A N/A Not Completed
 Pair D N/A N/A Not Completed

With cable unplugged on remote (user) end:

Interface Speed Local pair Pair length Remote pair Pair status
--------- ----- ---------- ------------------ ----------- --------------------
Gi1/0/1 auto Pair A 27 +/- 10 meters N/A Open
 Pair B 28 +/- 10 meters N/A Open
 Pair C 28 +/- 10 meters N/A Open
 Pair D 1 +/- 10 meters N/A Open

This led me to attach a cable tester on both ends, which showed that the last pair was bad. After I fixed that, now I get:

Interface Speed Local pair Pair length Remote pair Pair status
--------- ----- ---------- ------------------ ----------- --------------------
Gi1/0/1 1000M Pair A 48 +/- 10 meters Pair A Normal
 Pair B 48 +/- 10 meters Pair B Normal
 Pair C 48 +/- 10 meters Pair C Normal
 Pair D 48 +/- 10 meters Pair D Normal

Question is, why when connected @ 100mb, it shows Not Completed instead of telling me pair D is bad?

Leo Laohoo · ‎02-23-2017

Version 15.2(3r)E1

I think this is the ROMmon version. I don't see 15.2(3r)E1 (not with an "r") in the Cisco download site.

2960X (and 3850) has a "strange" way of implementing TDR. If there is a fault in any of the pair, the distance will be displayed. If there entire length has no fault, then the distance will be displayed as "0".

The "accuracy", however, is something else. Unlike the "older" 2960S, I recently found the accuracy of the 2960X TDR to be "questionable" at best.

My recommendation is to raise a Cisco TAC Case so that the developers may (a statement of probability) improve the TDR's function.

Don't ask me why, that's just how it was "implemented".

joebrug · ‎02-23-2017

Sorry, you're right, I posted the wrong thing:

Cisco IOS Software, C2960X Software (C2960X-UNIVERSALK9-M), Version 15.2(2)E3

markysharkey · ‎03-30-2018

I have to agree with Leo on the accuracy of the test on the 2960X.

On the 2960S results were usually pretty good. The 2960X results has just recently cost me a big chunk of personal reputation. For example, the below test is run on a 48 port 2960X running LANBase version 15.2(2)E5 with Bootloader 15.2(3r)E1. The cable was a straight-through 2 meter CAT5, the AP connected is a lightweight 1832i V3 and the local controller (not that it is relevant) is a 2504 with 8.5 firmware. Check the pair lengths and pair to pair readings

2960X-Gigabit#sho cable-diagnostics tdr int gig 1/0/2
TDR test last run on: March 30 18:21:15

Interface Speed Local pair Pair length        Remote pair Pair status
--------- ----- ---------- ------------------ ----------- --------------------
Gi1/0/2   1000M Pair A     28   +/- 10 meters Pair B      Normal
                Pair B     28   +/- 10 meters Pair A      Normal
                Pair C     28   +/- 10 meters Pair D      Normal
                Pair D     28   +/- 10 meters Pair C      Normal

That's really quite poor! To compare I ran the same test on an older Gigabit 2960S running LANBase 12.2(55)SE5 with bootloader 12.2(55r)SE

Lab-Sw-2960S#sho cable-diagnostics tdr int gig 1/0/2
TDR test last run on: March 01 00:06:59

Interface Speed Local pair Pair length        Remote pair Pair status
--------- ----- ---------- ------------------ ----------- --------------------
Gi1/0/2   1000M Pair A     2    +/- 0  meters Pair A      Normal
                Pair B     2    +/- 0  meters Pair B      Normal
                Pair C     2    +/- 0  meters Pair C      Normal
                Pair D     2    +/- 0  meters Pair D      Normal

That's much more like it. Ignore the date there is no NTP server configured for the 2960S.

What has gone so badly wrong with this test from 2960S to 2960X?

Leo Laohoo · ‎03-30-2018

Ho1y sh1t! That's one big, bad discrepancy.
My recommendation is to raise a TAC Case (make sure you involve your SE) and get them to investigate this as a bug.
Be warned that the Cat 2K is the "economy" version of the more expen$ive Cat3K and Cat9K. So please don't be disheartened when TAC or developers are/doesn't sound "interested" with this find.