public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
* what is the FS size limit for xfs_repair ?
@ 2008-10-03 15:04 Emmanuel Florac
  2008-10-03 15:23 ` James Braid
  0 siblings, 1 reply; 4+ messages in thread
From: Emmanuel Florac @ 2008-10-03 15:04 UTC (permalink / raw)
  To: xfs


After a system crash on a system with a 40TB XFS filesystem, xfs_repair
(32 bits binary, with a 64 bits 2.6.24 kernel) refused to work ("cannot
allocate block map, size xxx").
The 64 bits xfs_repair did the trick fine. 

I tried to reproduce the problem today by corrupting a 22TB filesystem.
However, 32 bits xfs_repair repaired the 22 TB FS just fine in the very
same configuration (64 bits kernel, etc).

I thought the limit would be at 16TB (32 bits limit for a filesystem)
but obviously this isn't the case. Is the limitation higher (like 32
TB) or is it a bug of etch 4.0r4 xfs_repair?

-- 
----------------------------------------
Emmanuel Florac     |   Intellique
----------------------------------------

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: what is the FS size limit for xfs_repair ?
  2008-10-03 15:04 what is the FS size limit for xfs_repair ? Emmanuel Florac
@ 2008-10-03 15:23 ` James Braid
  2008-10-03 20:39   ` Emmanuel Florac
  0 siblings, 1 reply; 4+ messages in thread
From: James Braid @ 2008-10-03 15:23 UTC (permalink / raw)
  To: Emmanuel Florac; +Cc: xfs

2008/10/3 Emmanuel Florac <eflorac@intellique.com>:
> I thought the limit would be at 16TB (32 bits limit for a filesystem)
> but obviously this isn't the case. Is the limitation higher (like 32
> TB) or is it a bug of etch 4.0r4 xfs_repair?

The limit is how much memory a 32-bit process can allocate - it's not
a limit of xfs_repair itself.

xfs_repair uses different amounts of memory depending on the
filesystem characteristics (number of inodes etc) - so a 32-bit
xfs_repair might work fine on an empty filesystem but not on a
filesystem with loads of files on it. e.g. we had a ~2TB filesystem
with millions of inodes and 32-bit xfs_repair failed.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: what is the FS size limit for xfs_repair ?
  2008-10-03 15:23 ` James Braid
@ 2008-10-03 20:39   ` Emmanuel Florac
  2008-10-04  0:22     ` Mark Goodwin
  0 siblings, 1 reply; 4+ messages in thread
From: Emmanuel Florac @ 2008-10-03 20:39 UTC (permalink / raw)
  To: James Braid; +Cc: xfs

Le Fri, 3 Oct 2008 16:23:41 +0100 vous écriviez:

> The limit is how much memory a 32-bit process can allocate - it's not
> a limit of xfs_repair itself.

OK, that makes sense. I'll keep my 64 bits xfs_repair at hand :)

-- 
--------------------------------------------------
Emmanuel Florac               www.intellique.com   
--------------------------------------------------

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: what is the FS size limit for xfs_repair ?
  2008-10-03 20:39   ` Emmanuel Florac
@ 2008-10-04  0:22     ` Mark Goodwin
  0 siblings, 0 replies; 4+ messages in thread
From: Mark Goodwin @ 2008-10-04  0:22 UTC (permalink / raw)
  To: Emmanuel Florac; +Cc: James Braid, xfs



Emmanuel Florac wrote:
> Le Fri, 3 Oct 2008 16:23:41 +0100 vous écriviez:
> 
>> The limit is how much memory a 32-bit process can allocate - it's not
>> a limit of xfs_repair itself.
> 
> OK, that makes sense. I'll keep my 64 bits xfs_repair at hand :)
> 

As a rule of thumb with a fairly recent xfs_repair, you need approx
128MB RAM per TB filesystem size (but it also depends on number of
inodes too).

Cheers
-- Mark

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2008-10-04  0:21 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-10-03 15:04 what is the FS size limit for xfs_repair ? Emmanuel Florac
2008-10-03 15:23 ` James Braid
2008-10-03 20:39   ` Emmanuel Florac
2008-10-04  0:22     ` Mark Goodwin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox