1

TLDR: Can a handful of problematic lines negatively impact the performance of the network as a whole? ... or would any problems stop once it reaches the switch?

Longer Version (background):

A school I support was recently modernized. They had professional cable installers run all new Cat 6 lines. Construction in general ran from April 2017 until roughly Dec 2017 - and the cabling was installed and certified in phases, with the last stage in October 2017.

In September (upon school resuming) we noticed they goofed by installing the patch panels in the back of the rack instead of the front. Due to the rack and room configuration, it was necessary for them to relocate the patch panels to the front of the rack.

Their installers came out and spent the day unpacking the cable bundles, and moving them to the front of the rack. Afterward they re-bundled the cables: from the back of the keystone jack it was secured to a cable support rod (which was attached to the 1U patch panel). The 1U patch panel's bundle was then bundled with its neighbors, fastened to the rack, which in turn became part of the 10-inch diameter bundle going into the ceiling (which is suspended by a cable hook before entering the cable trays)

We had a problem with one of our Unifi access points showing up as 100FDX instead of Gigabit. I did the usual troubleshooting. Despite having good continuity on all pairs, it was determined (through a process of elimination) that the problem had to be with the cabling / termination. The installers were contacted - they came out and tested - confirmed it was the line - repunched it .. and everything is working properly now.

When this problem first surfaced, I observed the keystone jacks from behind and saw that a number of them did not have the termination cap locked on - it has snapped loose during the cable relocation. After they re-punched the line, it was of course locked. I counted approximately 6-8 jacks that have their termination cap loose .. and another 10 that may have it loose (its hard to tell visually). I don't want to lay a finger on their cabling until after the warranty is up in a year.

If those lines were certified after each phase - the certification report demonstrated that the cabling and termination was to specification at the time the certification was run. But since they did not re-certify after moving the patch panels .. and we have now found a problem with a line that was previously certified .. wouldn't that call into question the entire certification report? Shouldn't they have recertified?

When the school corresponded with the network installer, the school was told that there was "absolutely no need to test cables" .. since it was only a single failure, and there have been no other issues with the system since the relocation of the patch panels ...

I know from experience that "not having issues" does not mean that there are not, or will not be problems. You can have full continuity on all 4 pairs .. and still fail certification. You will still be able to communicate on the line .. if the ethernet frames fail checksum and are discarded .. you'll get an increase in dropped packets with TCP .. and more Tx / Rx retries ... which in turn can slow down the communication on that line .. I'm guessing considerably.

But would the problem only be localized to that line - or could it affect the network as a whole? If so - would it be negligible on the whole, or a multiplicative effect?

If it is likely to be a multiplicative and substantial effect - do you have any links / resources you could refer me to, so I may read up further and reference it to the school?

I don't know if its needed to answer the question - but just in case .. we're using a 4 unit switch stack (functioning as one logical unit) .. Cisco SG500-52MP

Thanks in advance for your time and expertise ...

  • A cable that is re-terminated (on either or both ends) needs to be re-certified. I have seen things like having the blue on top of the blue-white at the termination cause it to fail Category-6 certification, but when the blue-white was on top of the blue, it passes. – Ron Maupin Jun 17 '18 at 0:25
3

A bad link will primarily just affect the communication between the two devices. However, there are some secondary effects that may come into play due to intermittent link loss or loss of important frames:

  • vital services through the link may become unreachable (DHCP, routing protocols, router discovery, ...)
  • spanning tree may play up with a flapping link, depending on how carefully your STP deployment was designed

You certainly don't want unreliable links in your network. Monitor each one for FCS errors, runts, or giants. Gigabit links are bidirectional on all pairs, so monitoring one direction should be enough. 10/100 Mbit links need to be checked in both directions.

When you've got cabling installed you need to make sure that it's up to spec. Getting certification for each link is essential, significant changes like moving the panels require recertification of the links. A simple check for continuity or shorts is a quick first diagnosis but passing that doesn't guarantee the cable to actually work at all.

  • To clarify for others: FCS errors = Frame Checksum .. means the Layer 2 Ethernet discarded the frame because it failed a checksum. Runts are packets that are smaller than the minimum size .. and Giants are packets that exceed the maximum size. – DHW Jun 22 '18 at 18:38
  • Turns out the school decided to accept the installer's word - we'll deal with problems "as they are discovered" .. I'll see if I'm able to rent a cable certifier from somewhere and do it myself ... or maybe I can just run transmission test and look for errors ... – DHW Jun 22 '18 at 18:40

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.