constant zfs data corruption

[Xin LI]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

JoaoBR wrote:
> On Friday 17 October 2008 15:49:54 Xin LI wrote:
[...]
>>> some idea?
>> For ZFS:  I think you need to remove these file.  ZFS scrub does not
>> recover file if you don't have redundant, e.g. copies=2 or raid-z, etc.
> 
> ok I see, I discovered this but should the machine crash when scrub reach the 
> damaged file?

I am just confused regarding to your issue.  If it's an on-disk
(physical) problem then I'd replace the disk.  If not, for instance, you
duplicated the contents of the disk to another and then do 'zpool scrub'
on it and discovered the errors, then we will face one of the two scenarios:

 - If you don't remove the file and access it, you will get EIO but no
data would be read.
 - If you remove the files then the place would eventually get overwritten.

Both case does not lead to a panic, but the first case *could* crash
your application if it is not well designed to handle I/O errors.

>> For the data corruption itself:  I think this is highly critical issue
>> if you are sure that the hardware is good.  What controller and disk are
>> you using?
> 
> disks are most seagate or maxtor sata-2 250G new, max 6 month old
> the Sata controller are mostly Nvidia as MCP51, MCP61, CK804

I have little experience with nVidia controllers since I almost use
Intel controllers exclusively...  Perhaps you can try installing
sysutils/smartmontools and see if there is some SMART errors?

Cheers,
- --
Xin LI <delphij at delphij.net>	http://www.delphij.net/
FreeBSD - The Power to Serve!
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (FreeBSD)

iEYEARECAAYFAkj8vycACgkQi+vbBBjt66BuegCeK/J0r56E1dhkL3SUpIKZ5kPH
TtMAniBBBSNmzitVipyEs5xvz8327BI6
=iUet
-----END PGP SIGNATURE-----