2

I got a bunch of 'used' servers, whose disks should have been checked, but they have been shipped around the county in crate which can't help. I just had one disk go bad (despite being mirrored, currently trying to get more details). The server was fine for about a week before everything ground to a halt this afternoon.

Is there any way 'accelerate' the failure of faulty disks, with the goal of bringing the disk to failure before we launch production services? Would doing lots of I/O with 'dd' or 'iozone' be a good way to test these potentially faulty disks? Any other tests/tools that would help recognized failures before they happen?

Dave M
  • 13,200
rainier
  • 21
  • 1
  • Related question: http://superuser.com/questions/84002/how-do-you-stress-test-your-hard-drives – sleske Mar 04 '11 at 00:01

1 Answers1

3

You might want to try running a DBAN on them. I might also try SpinRite, in "recovery" mode, since it does a lot of read/write on the disk.

Nate
  • 839
  • 4
  • 15
  • 29
  • It should be noted that DBAN has the side effect of wiping the disk. Just in case there's already something installed... – sleske Mar 04 '11 at 00:03
  • 1
    I think SpinRite would be the easiest way, assuming you have the time. :) – Hydaral Mar 04 '11 at 01:54