Kernel panics, need help interpreting

[Potatoshaped]

Hi all,

I have a headless instance of Debian Stable 10 (Buster) that crashes every few days. There are running 12 docker containers on it, and apart from that 1 Java Spring Boot application. The only other things I installed natively are certbot and some tools like iotop and such. So when I do a

Code: Select all

apt update

, usually there is nothing to update.

The previous time it crashed, I had a monitor attached and I could see a screen with a trace, to be found here: https://diederickdevries.net/goodies/20 ... 083956.jpg

Since it mentioned apic, I added 'noapic' to the grub boot line, but today it crashed again: https://diederickdevries.net/goodies/20 ... 45_HDR.jpg

The machine has the following hardware specs:

Gigabyte B450 I Aorus Pro Wifi (Mini-ITX) motherboard
AMD Ryzen 5 3600X CPU
Corsair Vengeance LPX 8GB DDR4 DIMM 2666 MHz (1x8GB)
MSI GeForce GT 710 2GB GPU (only because the CPU has no video)
Cooler Master MWE Gold 650 Full Modular PSU

Any help in interpreting these traces would be greatly appreciated.

[Head_on_a_Stick]

Try the kernel from the buster-backports repository (http://forums.debian.net/viewtopic.php?f=16&t=139003), third-generation Ryzen support is borderline with the stock version.

EDIT: and be sure to install the amd64-microcode package. It might even be worth downloading and installing the .deb package from sid, that has some extra fixes (but don't add the sid repositories, they will break your box).

[Potatoshaped]

Thanks a lot!

I really hope it works, but since these crashes only appear once every few days, I'll have to see. I installed the .deb using dkpg but why would the sid repo crash my system? Is it not considered stable?

Is there a way to find these kernel problems with newer CPUs in the logs before the machine crashes?

[Head_on_a_Stick]

why would the sid repo crash my system? Is it not considered stable?

Well sid is also known as unstable but that is a reference to the rate of change of the package versions rather than reliability (or lack thereof), contrary to popular misconception. However that package version flux means that new bugs may be introduced and also that mass package transitions may occur that will render some package uninstallable and some packages just plain broken for a time. If you add the sid repositories to your stable system then the package manager will preferentially install stuff from sid and also update your existing packages because the versions in sid are newer: https://wiki.debian.org/DontBreakDebian ... nkenDebian

Is there a way to find these kernel problems with newer CPUs in the logs before the machine crashes?

No idea, sorry.

[sunrat]

Is there a way to find these kernel problems with newer CPUs in the logs before the machine crashes?

Set up system journal and you will get logs you can read with journalctl. Just need to create /var/log/journal/ (as root) to enable it IIRC.

[Potatoshaped]

Well sid is also known as unstable but that is a reference to the rate of change of the package versions rather than reliability (or lack thereof), contrary to popular misconception. However that package version flux means that new bugs may be introduced and also that mass package transitions may occur that will render some package uninstallable and some packages just plain broken for a time. If you add the sid repositories to your stable system then the package manager will preferentially install stuff from sid and also update your existing packages because the versions in sid are newer: https://wiki.debian.org/DontBreakDebian ... nkenDebian

Fair enough.