My machine crashes frequently. It seizes up completely; mouse and keyboard are completely useless. When I check the syslog, I find nothing out of order prior to the crash. It just stops. The screen stays on just where it stopped.

Is there some way to monitor the hardware and log it?

I run Debian on an older box that uses a Duron 800 cpu on a MSI mobo.

If you're running X Windows and the system crashes as you state, the kernel dump will not be visible.

Exit out of X and leave your system at the console. When it crashes the kernel will dump to the screen. You will need to either photograph or write down the dump by hand.

Usually the output of the dump will lead you in the right direction of what is causing the problem.

I have an AMD 3200+ on a MSI motherboard and was having lockup problems due to SMP being compiled into the kernel. The kernel dump specifically stated that SMP was causing issues.

Scott

There are many reasons why the computer crash. One video module (driver) is corrupted or needs to be updated. Two power supply is going or power supply is too damn cheap and it is not able to handle the power wattage that the computer needs. Three memory corruption which can be minimized if ECC memory is used. ECC memory costs about 10 to 20 US dollars more than non-ECC memory which is not a lot compared to parity/ECC memory that costs twice as much. Use memtest86 to check the memory. Fourth, heat problems or dust bunnies has made a home in your computer. Use compressed air to clean your computer. DO NOT USE A VACUUM CLEANER OR ELSE THE STATIC CHARGE WILL DAMAGE YOUR COMPUTER. YOU HAVE BEEN WARNED.

Depending on the crash. If is an X11 crash, login through ssh and terminate X11. Then reload X11. The system should then be ok. If you do not have KVM switch and you yave magic key enabled in the kernel, you can push some keys to dump the memory and look as some logs.

Good suggestions. I have memtest86, but for the life of me, I can't figure out how to use it. All I get is "command not found."

You might look around in the security settings for an option to enable "Magic SysRq keys". This may enable you to access another virtual terminal and get out of X. Another option is to ssh into the box from another computer.

There are instructions in the kernel source: /usr/src/linux/Documentation/sysrq.txt

That is a good idea, but it doesn't help. I have been able to use the "Magic SysRq keys" only when the machine is healthy. When it seizes up, the keyboard is usless, including the SysRq key.

I finally figured out how to run memtest86. It is amazingly simple to do, but I'm appalled at how difficult it is to find directions.

Aside to dummies like me: memtest86 installs into your boot loader. To use it, re-boot; when you get to your Lilo or Grub screen, scroll down and choose memtest. That's it. Just let it run.

Ok, end of rant. I let memtest run until it stopped, about 4 hours. No problem.

Since I started leaving the machine at a console, it hasn't frozen. I have a littl Windowmaker applet that reads cpu temperature. How hot can a Duron 800 safely go? Or mobo?

If you have another machine, you can get serial null modem cable (usually about $5) and run a terminal program on there and see your console messages, even while X runs on the troublesome box.

Is there any way to log those console messages? Or, is there any way to stop my monitor from going dark? I turned off Xscreensaver, and I went through the CMOS and turned off anything that looked like a screensaver, but it still goes off. This morning it froze again, dark console, nothing to show for it, nothing to do but push the reset button.

Easier said than done. Our other box has Win 95 (non-negotiable). Assuming the connection, how does one access the Linux console?

You could use hyperterminal under windows to read the linux console. You just need to pass "console=ttyS0" as a kernel parameter. It will redirect /dev/console to the first serial port, which hyperterminal will be able to read.