ext3 fs errors 3T fs

[Damian Menscher]

On Thu, 19 Jan 2006, Andreas Dilger wrote:
> On Jan 18, 2006  16:16 -0800, Dennis Williams wrote:
>>
>> I have a ~3T ext3 filesystem on linux software raid that had been behaving
>> corectly for sometime.  Not to long ago it gave the following error after
>> trying to mount it:
>>
>> mount: wrong fs type, bad option, bad superblock on /dev/md0,
>>        or too many mounted file systems
>
> This sounds like the superblock has been overwritten.  There are occasional
> reports from > 2TB filesystem users of similar corruption.  It isn't clear
> if the problem exists in ext3 or if it is in the block or SCSI layer.
>
>> some more information on the system:
>> os flavor: Suse 9.1
>> kernel version: 2.6.5-7.202.7-default (various suse patches applied to
>>    2.6.5 kernel)

32bit or 64bit?

> RHEL4 (2.6.9) claims support for up to 8TB filesystems.  I don't know what
> patches they made, if any, in order to have this working.

FWIW, when we first tried using a >2TB filesystem on linux (I think it 
was FC3 at the time), we discovered filesystem corruption once data had 
been written past the 2TB mark on a 32-bit machine.  I'm guessing this 
is what you're seeing also.

We have been using (and filling) >2TB filesystems on 64-bit machines 
(FC4 and RHEL4) for some time now without problems.

Note that we didn't bother doing a detailed analysis of configurations, 
but rather tried a couple of variations until we found one that worked, 
so this could be a red herring (for those not familiar with the term, 
that means a clue that leads you in the wrong direction).

Damian Menscher
-- 
-=#| <menscher at uiuc.edu> www.uiuc.edu/~menscher/ Ofc:(650)253-2757 |#=-
-=#| The above opinions are not necessarily those of my employers. |#=-