Random connection drop on ethernet

HCHTech

Well-Known Member
Reaction score
3,848
Location
Pittsburgh, PA - USA
I have a solo-preneur architect that works from her home. She has a 4-year old Lenovo custom rig that started randomly (2 or 3 times per hour) dropping it's ethernet connection. It must have been from their gaming department, the case would hold at least a couple of normal-sized towers. It's got a 128GB SSD for the OS and a spinning disk for storage. Windows 10, but originally came with 8. I don't know if Win10 was an upgrade or fresh install (probably an upgrade). The machine is currently on the fall creators update (1709) but I couldn't get confirmation that the problem started with that update.

It's an old house, but someone wired a jack in the wall where the computer is that goes to a makeshift wiring closet on the floor below. There is a FIOS modem and a 16-port cisco switch - the little unmanaged one.

Troubleshooting remotely first, I had her plug a laptop into the same cable - it ran fine for hours, so the problem is likely with the computer, not the infrastructure. I book an onsite appointment and find that all of the equipment looks good and the jack wiring is solid. I replace the patch cable just because, then take a look at the computer. The connection is good and web pages seem to work ok. I get about 80mbs up and down over 3 different speed test sites. I run a constant ping to the router (no drops over 20 minutes, then a constant ping to an external site, no drops for about 15 minutes.

Then the connection drops. I'm connected to the router, but not to the internet. The laptop in the house still has internet. I reset the TCP/IP stack and the connection wakes up again. I go down to the wiring closet and plug the connection directly to the FIOS modem, taking the switch out of the loop. Back upstairs I get the same behavior. It works well for about 20 minutes and then the connection drops.

I plug my laptop into the wall jack used by the computer and everything is good - no drops.

As a test, I decide to replace the network card. It looked like that solved the problem, but the next day the phone rings and the problem is back.

I go onsite again and this time I look at the system event log, and find a single error at the point the connection drops. DNS Client event 1014. Name resolution for the name XXXXXX timed out after none of the configured DNS servers responded. First, I tried setting the connection to use Google DNS servers, but that didn't solve the problem, I still got a dropped connection with that error after several minutes.

So let's dig into that error - it's not one of the normal ones. Some Googling takes me to this technet article that recommends disabling IPv6 and disabling RSS, Autotuning & taskoffload with netsh commands. Ok, simple enough.

I also find a few threads of unhappy people with the same problem. Maybe related to NICs using the Realtek chipset. Both the original NIC in my customer's machine AND the replacement NIC I installed are Realtek, so I can't prove or disprove that link. Some folks had better luck with older drivers, and some had luck with disabling power managment on the NIC.

So in addition to the netsh commands, I disabled power managment of the NIC and it worked for the 40 minutes I was still onsite. I think if the problem returns, I'll put in an Intel card. Gotta love driver problems. No real questions here, I guess, just expressing some frustration. One of the risks of upgrading old hardware to Win10 is obnoxious, tail-chasing errors like this. I hope this might help someone else.
 
Probably not the issue in your case - but I had an intermittent network issue just like this last week. Similar thing that it would work for 45 minutes, maybe a few hours, then all of a sudden no internet. Could still ping the router (well, the IP anyway. read on...). Their iMac worked using the same cable/wall jack, so not a cabling issue. Tried setting static IP/DNS, resetting network stack, drivers etc. No luck

By chance they mentioned someone had recently replaced the switch (residential client. Only a 5 port dumb switch). Apparently they were in a rush and just bought the first switch they seen. Turns out it was a smart switch with default IP of 192.168.1.1 - the same as their router.

Still not sure why only one PC was effected, but it was definitely the switch at fault. Since changing the IP they haven't had any issues.
 
Back
Top