Why does my Ubuntu 18.04 randomly shut down?

Question

I'm running Ubuntu 18.04 (start with server edition, because the hardware is pretty weak) on some older hardware - a desktop. Memtest ran fine on it before I installed, but it seems that it will randomly shut down - typically when I'm doing something like a sudo apt-get update. It appears to go through the normal shutdown process - I see all the shutdown messages, etc. But it's definitely a forcible sort of stop since I have to run dpkg --configure -a or whatever it is if I was in the middle of installing software.

Any clue what could be happening here? I have a theory it might be related to power consumption but I'm not sure why that would cause what I'm seeing. If I run glxgears I can run like 10 of them at a time (slows my FPS way down, of course) for a couple of minutes and then beeeooooop! Off my machine goes.

Is there somewhere I could look?

Solution 1

In my case, it turned out to be an Nvidia driver that caused random shutdowns. I had the driver 396.54, but removing it and installing the 390.87 fixed the problem. You can do something like

sudo apt purge 'nvidia*'
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt install nvidia-390

Solution 2

For me it was also the temperature limit, although I think it is a false postive and something makes that value jump somehow. I recently changed batteries on my old Lenovo T430s and these things started happening.

This way I found out:

cat /var/log/kern.log | grep shutting

And it showed this:

Jan 10 11:36:34 kernel: [166814.992608] thermal thermal_zone0: critical temperature reached (128 C), shutting down

Solution 3

If the above doesn't work for you it might be because the current drivers might not be recommended for Ubuntu 18.04 on your machine.

First remove all current drivers

sudo apt purge 'nvidia*'
sudo add-apt-repository ppa:graphics-drivers/ppa

Try running

ubuntu-drivers devices

and install the driver that has recommended next to it using

sudo ubuntu-drivers autoinstall

then reboot

Comments

[RobotHumans]

I had a similar problem. It's probably a heat sensor (just spitballing, I haven't checked). Where if I run a VM within virtualbox on a single core machine, it has no problem pegging out the processor at 100%. So, heat most probably is the issue if it sits there pegged. Windows probably just had more loosey-goosey heat tolerances or doesn't do heat based shutdowns.

[Wayne Werner]

@RobotHumans how would I be able to check that?

[user535733]

Can you please tell us more about the shutdown messages that you see? Also, review /var/log/syslog at the time of shutdown for helpful warnings and notifications. If you run across any, please add them to your question.

[Rinzwind]

If it is truly random it is likely a hardware error. Check in /var/log/ if you can find an log that got updated and if it shows any errors.

[Wayne Werner]

I couldn't find anything in /var/log/syslog - not even anything about shutting down. I even did an ag power (grep -Inri power . equivalent) in /var/log/ with nothing suspicious. Any other terms I should keep an eye out for?

[RobotHumans]

I'ld also look at dmesg output. If you see kernel panic anywhere, that's your problem.

[Chai T. Rex]

There's a program called stress that can stress the CPU, drives, or memory to see if stressing one causes a shutdown.

[Wayne Werner]

Finally got around to reinstalling and this appears to be it

[Ben]

So how did you fix it?