A few months ago, I found a failing NIC that knocked the network down. Before that, one of my ISP-provided boundary devices started failing. Yesterday, I lost the Ethernet side of my network. That is, the Wi-Fi from the boundary device kept working while the rest of the network crashed. Fortunately, I had a pretty good idea that my primary GbE switch might need replacement and had already ordered one through Dell in July. Even more fortunately, a quick switch replacement fixes network woes, and brings Ethernet back to life.

Literal Switch Replacement Fixes Network Woes

The funny thing is, I’ve been using the same switch in my office since we moved into this house in April 2006. And when I went to re-order, the same switch remains available at a knock-out price of US$40. It’s the venerable Netgear GS-108 unmanaged 8-port GbE switch and it works like a charm. I guess 15-plus years of uninterrupted, heavy-duty service ain’t bad. In fact, I’ve used all 8 ports all the time and that device is as close to a network backbone as the 12-15 devices around our house can access.

The blurb on the NetGear site reads “Set it and forget it, energy-efficient switches are built like tanks and last for decades.” In fact, I can’t remember when I bought the original. I know it must’ve been some time around 1998, when I moved into my previous house. Thus, I’d have to agree with that seeming hyperbole.

Dead-Simple Replacement

I unplugged power jack from the old switch. Then I removed all 8 of the RJ-45 cables plugged into its face (see lead-in graphic). I unpacked the new device, plugged in its power supply, and plugged in the RJ-45 cables. The power light came up, after which the activity LEDs started blinking. Problem solved.

There’s another GS-108 of about the same vintage upstairs under my wife’s desk, where it serves to distribute Ethernet to that floor of the house. I have another replacement in my spares closet, ready to take over for the old one should it fail, too.

How I Knew It Was the Switch, Stupid!

When the Ethernet side of things goes down, it has to be a device that makes the side work. That means it could have been a switch, of which I have 4 on my network. One is in the recently-replaced router/wi-fi/switch device from Spectrum, replaced in June. Another is in my Asus AX6000 wi-fi/switch/router: it’s Wi-Fi was still working so I guessed that meant the switch portion was still working, too. Thus, it was likely one of the two GS-108s. Logic dictated the heavily-used one in my office would be the one to fail first. This time, logic prevailed — or so it seemed.

i’m just glad I had a spare on hand. I’m even gladder that the switch  swap was as simple and painless as I hoped it would be. Sometimes, here in Windows-World one does catch a break. With plenty of real work to do yesterday, I was appropriately grateful.

Or Maybe Not, But Real Cause Emerges Quickly

About two hours after I posted this, my problems returned in full force. That left only one other possible cause: some element in the Ethernet network had to be failing intermittently. I had two prime candidates:

1. My 8-year-old Surface Pro 3 dock, whose GbE port has been flaky in the past. That wasn’t it.
2. The cable from my switch to the filing cabinet by the window in my office goes under my desk, where I can’t help but kick that cable occasionally. Apparently, I’ve kicked it often enough to introduce an intermittent short. Now that it’s removed from the network all is once again good.

I guess I can keep my ancient GS-108 Switch around as a spare, because it obviously was NOT the cause. And that’s how it goes when troubleshooting intermittent Ethernet gotchas. Live and learn!

My first real networking job was as a Networking Consultant for Excelan in 1988. That company was purchased in 1989 by Novell, where I stayed quite happily until 1994. My initial training for the position included learning a hardware-based protocol analyzer (the LANalyzer, in fact). One of the things we learned in class was a coax-based 802.1 10 Mbps transceiver could crash an entire physical LAN. This device had a classy alias: “vampire tap.”  It was scre-clamped onto a thickwire coax cable to add one or more  network ports. Sometimes, its built-in circuitry would go bananas and overrun the network with bogus traffic. This problem, known as a jabbering transceiver error rears its ugly head recently. It happened  on  one of the Chez Tittel GbE switch domains.

When Jabbering Transceiver Error Rears Its Ugly Head, Divide and Conquer

Here’s a quote from the 2000 classic by Charles Spurgeon: Ethernet: The Definitive Guide

The quote comes courtesy of Google books, pg. 107.
(I still have a hardcopy on my bookshelf).

I’m pretty sure that NICs don’t have transceivers any more, so they aren’t really subject to such failures. But similar behavior — specifically, failure of a switch domain — is well-known to occur when hardware problems bedevil a LAN segment. For a while there, I was chasing random network failures in my office. They would kick all the machines off the switch, but would gradually let everybody back on.

It wasn’t until I quit using the built-in GbE port on my retiring X220 Tablet PC that the problems stopped. I was able to confirm the issue by plugging the RJ-45 cable back into that until and watching the circus start back up. If I switched to a USB dongle instead, the GbE domain attached to either or both switches in my office worked fine. One is a standalone NetGear 8-port GbE switch, the other an 8-port switch integrated into my Asus 802.11AX WAP/router.

Historical Note

Divide and conquer was the recommended troubleshooting method to identify a jabbering transceiver. One would subdivide the cable segment by interrupting it at a repeater, and terminating each sub-segment. Whichever segment stayed broken had the failing device. Repeat until the device can be identified, then replace it. I did this for TRW in Austin in 1988 on an actual service call there…

It wasn’t really until I started the trip down memory lane to my first-ever Ethernet networking class in 1987, and my trip to TRW,  that I understood what was happening. The built-in GbE interface was failing, and acting like a jabbering transceiver. I can’t exactly say “everything old is new again.” But I can say, an old lesson learned came in handy. And indeed, that is the way things sometimes go, shooting trouble here in Windows-World!