[RESOLVED] Ethernet randomly stops working after upgrading from 22.04

[jasoncollege24]

network is a 1 gigabit connection. Worked fine in 22.04. Every so often (usually once or twice a day) the network stops sending/receiving data. If I click "Disconnect" on the connection, it vanishes for a few seconds then appears again. clicking "Connect" at this point restores the network.

CPU is a i7-4790 on a Z97-DH3 motherboard. For other reasons, I've recently replaced my router, and several cables including this one. I've tried restarts, power resets, and apt updates to no avail. I have another system running the 24.04 (clean install) which doesn't have any ethernet issues. That system dual-boots with Windows, which also has no ethernet issues.

This machine is remotely administrated, and somewhat tedious to get to physically. it runs a cloud service, and a smart home service. I'd like to try to keep this system online as much as possible, without having to physically get to it. If anyone can help, I'd greatly appreciate it. Need more information? Let me know.

[jlittle]

I have a Gigabyte motherboard of a similar vintage. It has an Intel ethernet chip that uses the e1000e driver, and unpredictably starts exhibiting the behaviour you describe.

Unfortunately, on my hardware there is a bug in the hardware, so a driver update can't fix it. But, if I tickle the connection, the problem occurs much less. I run a script that pings my router every 4 seconds, and writes to a log file. If I suspect a connection problem, I look at the end of the log file to see if the regular 4 s ping appears.

[jasoncollege24]

Thanks for the reply.

With a little help from ChatGPT, I discovered that I'm using the same e1000e driver. I told it to search the web for a potential solution, and it suggested that I disable offloading to the NIC by adding the service file shown in the code block.

Unfortunately this didn't resolve it. I'm not good at troubleshooting network issues in Linux, because I rarely have them. Usually, the network just works. I could probably do something similar to what you're doing... Have a script ping the router every 15 seconds, and restart the NetworkManager service if a ping fails.

[Unit]
Description=Disable EEE on Intel e1000e NIC
After=network.target

[Service]
Type=oneshot
ExecStart=/usr/sbin/ethtool --set-eee eno1 eee off

[Install]
WantedBy=multi-user.target

[jlittle]

When I was troubleshooting this problem in 2017 or 2018, (all this from a 5 year old memory:

• It more commonly affected servers.
• The cause is in the ethernet chip itself, and cannot be fixed with settings.
• An updated driver from Intel helped, and made reconnecting faster. I downloaded and built a kernel module for a year or two,. The module had to be rebuilt for every new kernel, a pain, and eventually I set up dkms to do it. But then, I decided that the Linux version of the driver had been revised; it was hard to tell because the version numbering changed, and became useless.
• To ameliorate the problem, I run a ping command, that pings my router every 4 seconds:

    ping -i 4 192.168.1.1 >> ~/log/ping.log

When the ping is running, the problem is much less, sometimes occurring weeks apart, sometimes a few times in a day. If I suspect a problem, I tail the log file to see if the ping is working, and if not I disconnect, wait a few seconds, then reconnect.

HTH

[jasoncollege24]

I disabled the service that I made, then repaired my cloud server VM. It seems as though it has resolved itself at least for now. The system is being used as a server, but it using the non-server version of the OS. It has been stable for the last few days, so I'm hoping it has fixed itself. I'll update in a few more days on whether the issue is actually resolved.

[jasoncollege24]

Sorry for taking so long to post again. The above changes did not resolve my issue. Instead, I added a new PCI-E ethernet card to the system. Since I started using that card for my network, I haven't had a single disconnect.

For anyone else that comes across this problem, adding a new network card is probably your cheapest solution.