HDD Stress Test with Python

[Bart Nessux]

A bit off topic, but here goes: Does this look like a reasonable hard
drive stress-test? I run it repeatedly for a few days on our new
PowerMac G5's (they come with Python 2.3 installed!!!) which we've had
drive problems with. Two 80GB SATA Maxtor drives replaced already. Only
been in use for a few days. Apple is replacing them with Seagate drives,
not that that matters. Here's the script:

def dd_test():
import os
import time
print
print "DDing first partition to second partition... please wait."
fp = os.popen("/bin/dd if=/dev/disk0s3 of=/dev/disk0s5", "r")
fp
fp.close()
print "DD has completed."
print
print "Formatting first partition UFS"
print
fp1 = os.popen("/sbin/newfs /dev/disk0s3", "r")
fp1
fp1.close()
print
print "Formatting second partition UFS"
print
fp2 = os.popen("/sbin/newfs /dev/disk0s5", "r")
fp2
fp2.close()
print "Sleeping 60 seconds before starting again..."
print
time.sleep(60)
dd_test()
dd_test()

# Make two equal sized partitions on Mac internal HDD.
# Unmount the internal HDD, boot from firewire drive
# and run script.
# Used 'pdisk /dev/disk0 -dump' to get info about drive.
# Used 'newfs /dev/disk0s3' to format a partition UFS.

[Skip Montanaro]

Bart> A bit off topic, but here goes: Does this look like a reasonable
Bart> hard drive stress-test?

...

Bart> fp1 = os.popen("/sbin/newfs /dev/disk0s3", "r")
...
Bart> fp2 = os.popen("/sbin/newfs /dev/disk0s5", "r")
...

My guess is that newfs isn't all that random in its behavior, nor that it
touches all that much of the disk. I think it mostly lays down inode
structure. I doubt it even zeroes out the data chunks. You might want to
do something like build multiple copies of some huge program (Emacs? GCC?
Gimp? Darwin kernel?) in parallel, while constantly creating and deleting a
bunch of large files stuffed with random bits.

Skip