Wan speeds slow to a crawl must reboot to resolve

My Pc is wired and gets 900mbps consistent for a couple days then magically drops to about 10-12mbps. Its only happening on this machine while all others on my network stay 900+. I have tried a few things my LLMs recommend and they find nothing in the logs when the speed drop occurs. Nothing downloading or uploading. Only a reset of the machine brings the speeds back to normal.

there is only 1 thing in the logs between the speed tests

That 10-12 Mbps number jumps out cuz that’s basically the ceiling of a 10 Mbps link, so my guess is your Ethernet port is renegotiating down to 10 Mbps after a while, not that anything is actually throttling you. As far as I know umbrelOS doesn’t do any bandwidth limiting on its own, and the fact that a reboot fixes it also fits a link/driver thing.

Next time it slows down, before rebooting, open Settings > Advanced settings > Terminal > umbrelOS and try:

cat /sys/class/net/*/speed
sudo dmesg -T | grep -iE “link (is )?(up|down)|r8125|r8169|igc|e1000|timed out”

If the first one shows 10 (or 100) for your Ethernet interface instead of 1000/2500, that’s your answer. The second one will show if the port has been dropping and coming back up.

Also worth trying: just unplug the Ethernet cable for 10 seconds and plug it back in, then rerun the speed test. If that fixes it without a reboot, I’d swap the cable and try a different port on your router/switch. I’ve seen flaky cables and certain switch ports cause exactly this.

What hardware are you running it on? If it’s a DIY box, lspci | grep -i ethernet will show which network chip you have, some Realtek and Intel 2.5G chips are known to be a bit finicky.

Its on a dell optiplex 7060. I have moved cables around without a reboot and my switch still showed 1gb lights when it happens, but im moving to all new Unify switches and router over the weekend so I hope the issue goes away. I Had hermes agent checking stuff too so im pretty sure it thought the Nic was till at 1gb. I just wanted to toss it in here in case by chance any issue like this was happening to anyone else. I’ll report in my resolution if I find it.

Ah nice, hopefully the new UniFi gear just sorts it. Going back over your screenshot though, I noticed only the download tanks when it happens. Ping stays around 16 and upload sits at ~41 the whole time. If the port was actually falling back to 10 or 100 Mbps, upload would get dragged down too, so I think my link speed guess was off, which also fits with your switch still showing 1G.

A download-only slowdown that goes away after a reboot makes me think more of something on the box itself, either the network card/driver or the system getting bogged down. The 7060 has the onboard Intel I219-LM (e1000e driver), and that one is known to act up on Linux sometimes.

If it comes back after the switch swap, I’d grab these before rebooting (Settings > Advanced settings > Terminal > umbrelOS):

sudo dmesg -T | grep -i e1000e
curl -o /dev/null “https://speed.cloudflare.com/__down?bytes=500000000”

The first shows if the driver logged any hangs or resets. The second downloads a 500 MB test file straight from umbrelOS, outside of any app. The Average Dload column is in MB/s, so around 110M is roughly gigabit. If that’s fast while MySpeed is slow, it’s more likely something on the app/Docker side than the network card.

Might also be worth peeking at Live Usage when it’s slow to see if CPU or RAM is maxed out. Your log had a big backup job and an app update right around when it dropped, could be a coincidence but worth keeping an eye on.

Curious what you find either way!