I have assembled my desktop PC about 2 years ago. It’s fairly beefy (AMD Ryzen 9 3950X 16-Core Processor, 128Go RAM, nVidia RTX 3080 Ti). It’s running debian stable.
Once in a while (not that often, but like every 2 weeks or so), seemingly at random times, not especially under heavy loads, the system crash and freeze, irresponsive to even the linux sysrq magic keys. I never manage to find what was the cause. One interesting fact is that when it happens, for some reason it seems to “freeze my network” too, ie, other (ethernet) devices on my local network have no connectivity anymore. They’re all connected to the same router, but not through this crashing PC. Connectivity comes back as soon as I force shutdown the crashing PC.
What can cause this and how could I fix these freezes?
Your network card is hammering the network with packets and not taking a break. It doesn’t give the rest of they network a chance to talk.
If this was a windows machine, I would start by reinstalling the network driver, but I don’t know Linux well enough to say.
Oh this would explain why it kills the connectivity of all ethernet-connected devices. The ethernet interface is the one on the mobo. Drivers are included by the linux kernel AFAIK. The problem persisted across 2 debian versions so I am not sure re-installing drivers would do anything here. But thanks for the plausible explanation about the network issue!
You won’t have much luck with doing anything to the driver part of it, but you could try a custom kernel. There’s two advantages to that, one is it would be more recent than whatever kernel that Debian is using, and the second is the optimized networking stack, which speeds up processing of packets and improves the congestion handling algorithm. I’d recommend the Xanmod kernel for this: https://xanmod.org/
Alternatively, if we suspect your network is the culprit then the solution could be as simple a buying a new card and disabling the builtin one.
I like my debian vanilla but thanks for the suggestion. The other network card would be interesting to try out. I don’t really suspect the network card, since I have no idea whether the network block is a consequence or a cause here.