System freezes. Can't find anything in the logs

Question

My linux desktop system freezes (aptosid/debian sid) sometimes and after a reboot I can't find any infos about the reason for the freeze in messages/dmesg/syslog/Xorg.*.log.

When it "freezes" I can still move the mouse in X and sometimes even move the windows for a short time before the system stops responding. Last time I could even switch to the text-terminal and enter my login name, but every text-console stopped after that and displayed no password prompt.

SSH login also does not work after the freeze.

I think the problem started some weeks (maybe 1-1.5 months) ago.

I have a windows partition (win7 64bit) on the same machine which I use mainly for gaming (in the last time mostly StarCraft2). I can't rember any crashes there in hours of playing.

I haven't found a way to force the freeze yet which makes debugging not easier.

Solution

Sounds like a potential hard drive problem. The behavior you describe is exactly what you would see if the hard drive suddenly stopped responding to requests, or started returning errors. Try running smartctl -a /dev/sda, where sda is the device of your hard drive, to see if the hard drive has logged any errors.

You would see an explanation of the error in the kernel's dmesg output. Of course, if you can't login, you can't run dmesg. However, if you log in first and run the command on the text console, it may still be in memory so that you could run it again after the system has crashed.

The other option would be to use the netconsole feature of Linux to send the debugging information directly to another machine instead of writing it to disk. This is a little tricky to set up and requires two machines, but is a good thing to try if everything else fails.

Comments

[gadu]

This seems possible. The bad thing is that my main disk is a relativly new 256GB SSD, and i don't think "some unclear freezes under linux" will be good enough to get an replacement. Smartctl says No Errors Logged but i will try it with netconsole.

[cheshirecatalyst]

SSDs are crap. Then again, hard drives are crap too, so you pick your poison. :) Just as a random example, here Andrew (bunnie) Huang laments the same types of problems. My suggestion? Go with RAID if you need reliability, otherwise just keep a close eye on things and don't be too surprised when hard drives or SSDs fail.