Broadcom Netxtreme Gigabit Ethernet

[Smacka Shock]

Wehave the weirdest network issues with a few Dell servers, all at different customers, where the Broadcom NetXtreme NICs randomly disconnect from the network until the NIC is disabled and re-enabled.

The servers will run for about a day without any problems, and then the traffic on the NICs start to drop. No errors or anything are logged in the eventviewer and no Windows updates have been installed recently that can cause anything to break. I have had permanent pings running to 8.8.8.8 and our internal gateway and when the problem occurs, everything just starts to timeout. I then need to either physically unplug the network cable and plug it back in or I need to disable the NIC inside Windows and re-enable it. Doing so will immediately restore all connectivity for the time being, until the next time it decides to go down again. Whenever the problem happens, the activity LEDs on both the switch and server NIC port are on and flashing. Nothing that indicates there is a physical connectivity issue.

This has happened with multiple customers and only on Dell servers. Rebooting the server, installing updates and installing latest NIC drivers do not resolve the issue. Has anyone ever seen this before?

Turned out the problem was having VM Queues enabled, supposedly the issue was resolved in a firmware update or driver update but I never saw the difference. Same with disabling it computer management. Best solution was buying Intel cards instead.

It sounds like a similar problem indeed. When we get the first alerts from our monitoring system, pings still go through but with about 2000ms delay until eventually the traffic goes completely dead and time out. Then all I do is restart the NIC inside Windows and everything is restored. But this is causing a lot of problems with running backup jobs and other systems that rely on a stable connection.

I was going to try the USB to ethernet adapter option as it is the most inexpensive solution. What was your experience with it? Did it resolve the problem and is it stable enough to run on a production server?

We added the USB NIC and compared to the broadcom it was fantastic. The toughest part was explaining why a $30 adapter was out performing a brand new much more expensive one. But we had it running sql, file shares, active directory without an issue.

Thank you. I will pick one up today and install it tomorrow and see how it performs over the weekend. Will mark as correct answer if no issues arise. Right now the server is going down about twice a day.

I picked up this server for like $20 and just using it for my homelab setup. The problem is that the ethernet speeds are peaking at 8MB/s on the LAN and even worse to the internet down to like 5MB/s. This would normally point to being connected at 100Mbps but both server and router are reporting a 1000Mbps link. This is both pulling to server and pushing out which rules out drive speeds. Plus testing on SSD. My computer will do gigabit speeds connected to same router. I have used the same cable on the server with no change. Running Proxmox on Debian with kernel 5.3.18-2-pve. Any ideas or suggestions would be greatly appreciated

3a8082e126