[lug] CPU tester, Disk tester

Andrew J Jenkins andrew at winston.homelinux.org
Wed Jul 2 17:13:44 MDT 2003

I just got through a periodic hard disk disaster, don't know if any of 
this will help:

hdparm can be used to test hard disks or spin them down to reduce power 
consumption, if you think it's a power problem.  I'm guessing that since 
makewhatis crashes, it's probably related to the disks - maybe when the 
hard disks are really going, your power supply can't power the 
motherboard, or something similar.

"hdparm -tT /dev/hda" will do some speed testing on your hard disks. 
 Something like "hdparm -S10 /dev/hda" will shut your hard disk down 
after 10 seconds, although lots of filesystems (ext2/3?) periodically 
write data to the disk that might cause it to continually stay spun-up, 
or spin up and down constantly.

"badblocks -n" (in my e2fsprogs version 1.29) will do non-destructive 
read and write testing to all the blocks on a device that is not 
mounted.  I think this is a relatively new feature of e2fsprogs - I'd 
double check the man page for your version to make sure I'm not telling 
you to do something destructive on older versions.

The 2.4 and 2.5 kernels have pretty good environment reporting.  Try 
"watch cat /proc/acpi/thermal/0/status" - for me, this displays CPU 
temperature in decikelvin.  If it's higher than 70C (3430 dK), then you 
might have cause for concern, adding more cooling.  You could do this in 
one terminal, and then build a kernel in another - see how far it 
spikes.  My motherboard shuts off if the CPU (athlon-xp) gets to 70C.

More information about the LUG mailing list