On 4/12/07, Les Mikesell lesmikesell@gmail.com wrote:

If the drive supports SMART, smartctl should be able to test it and display the diagnostics. I'd try to get a replacement before the warranty expires.

But of course!

What's interesting is that I currently have the SMART in the BIOS disabled, and the disk says it doesn't support SMART capability (per smartctl). I don't know if that's because the BIOS SMART is disabled, but I'm still going to put some pressure on it.

Thanks.

mhr

Hi,

To exercise the disk you could use a tool like iozone3. It's intended to give you a broad overview of io/disk performance. Obviously you could use badblocks to read/write every sector of the disk which will exercise it quite well, however with both tools I would advise to make sure you have a backup of stuff first.

Also, iozone is not intended to do anything related to disk errors or other io failures; it might just uncover when/how and error occurs coincedentally. With badblocks, it might show io errors at a certain part of the disk.

I've had SMART errors on disks that would still work for about a year after the first warning came up, however it is usually not a good sign for long term operation of the device(tm).

Grtz,

Rubin.

Mark Hull-Richter wrote:

...

I recently added a Seagate 400Gb SATA drive to my system, and it has been behaving strangely since I put it in. for one thing, the BIOS S.M.A.R.T. came up with a warning the last time I booted with it enabled, saying that I should backup my data and replace the disk (!).

I still have not made any irreversible data transfers to this drive, and I have some time yet to take it back, but I'd like to know for sure that it needs it, or at least have some reasonable evidence of failure.

What is a good program out there that exercises a disk to give some assurance of errors or lack thereof?

I found that older versions of smartmontools (even the one included with CentOS 5) do not handle newer onboard SataII controllers. In my case, one system is running CentOS 4 -- that motherboard has an Nvidia MCP northbridge (sata_nv). I ended up building kernel 2.6.20.1 to get it working better. My other machine has an Intel P965 Express northbridge. In the former case I downloaded the smartmontools-5.37.2 source and built it; in the latter I downloaded the fedora core 7 development source rpm for smartmontools (also 5.37-2). In both cases smartctl -a started working properly with the combo a newer kernel and latest version of smartmontools. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos

Now I'm wondering what's the best way to destroy all the data on the drive so I can return it without my stuff, unreliable or not, on it.

you can use shred and even loop it a few times

http://www.die.net/doc/linux/man/man1/shred.1.html

On 4/13/07, chrism@imntv.com chrism@imntv.com wrote:

Also, doesn't the seagate utility have a low level format function? I haven't used one of those utility disks in ages, but that used to be standard fare.

It did and there was, but the SeaTools does not include such an option, and the Disc Wizard, or whatever the heck it is called, doesn't run if there's a non-MS MBR on the disk (as in a Linux/Grub MBR), so shred and/or badblocks will have to do.

Thanks.

mhr

On Fri, 13 Apr 2007, Mark Hull-Richter wrote:

Capstone (so to speak):

I booted from the Seagate CD and ran both the quick and full diagnostics on the disk. It failed them both.

Now I'm wondering what's the best way to destroy all the data on the drive so I can return it without my stuff, unreliable or not, on it.

I suppose a dd from /dev/zero to the whole disk might work - am I right? I'm not sure what block size or how many - it's a 400Gb drive....

Thanks.

Mark Hull-Richter, Linux Kernel Engineer

Or even better badblocks -w

I would add -v and -p some_number See man page.

Some of your data can be left in reallocated sectors.

Wojtek

fredex wrote:

I'd guess that several passes with data from /dev/random might be a good thing.

I've tried that, tis kinda brutally slow. dd if=/dev/random of=/dev/sdb bs=1048576

Ok... ok...

I will take on the job of *exercising* your disks for you if you pay me well, and buy a new pooper scooper and leashes for the long walks I will have to take them on....

:-)

- rh

-- Abba Communications Internet PO Box 7175 Spokane, WA 99207-7175 www.abbacomm.net

Mark Hull-Richter wrote:

Capstone (so to speak):

I booted from the Seagate CD and ran both the quick and full diagnostics on the disk. It failed them both.

Now I'm wondering what's the best way to destroy all the data on the drive so I can return it without my stuff, unreliable or not, on it.

I suppose a dd from /dev/zero to the whole disk might work - am I right? I'm not sure what block size or how many - it's a 400Gb drive....

the block size doesn't matter except for performance. use 1048576 (1MB) or something.

Alternately, run the seatools destructive sequential write w/ random data diagnostic, at least 2 full passes of random data will completely scramble any residual data beyond any possibility of recovery. the old DOD standard 1111, 1010, 1100, 0011, 0101, 0000 thing is obsolete, and was based on MFM/ESDI/PLL bit encoding patterns, which no longer apply to modern disks...

but, if the drives failed the diagnostics, it may not even let you write patterns over the full media.

I suppose you might explain your data destruction requirements (are these corporate policy, or government/military based, or just personal policy?) to the Seagate customer services people, and ask for their guidance. surely, they have a provision for this as it can't be uncommon... they probably won't let you physically destroy the drive, as I've heard something like 80% of drive returns are retested and turn out to be 100% AOK, and can be reformatted, 'refurbished', and put back in service.