On 4 Mar 2003, Stephen C. Tweedie wrote:
> Hi,
>
> On Mon, 2003-03-03 at 23:50, Andy Vaught wrote:
>
> > We've been observing some consistent filesystem corruption under ext3.
> > The problem is that a directory created by one of our applications
> > vanishes for some reason. The directory is created on installation the
> > application. The files it contains are written infrequently (days to
> > months) and read as part of every user command. The directory is under
> > /etc and always has the same path.
> >
> > So far, we have had three separate customers with this problem and we
> > have observed it once ourselves. One of our customers was able to use the
> > ext2 filesystem editor and reattach the orphaned inode, recovering his
> > data. When we observed it here, fsck -f (ext2 and ext3) didn't find
> > anything wrong with the filesystem. We are running Debian woody (stable)
> > with a 2.4.18 kernel.
> >
> > We haven't had reports of other files vanishing, just this one directory
> > and only under ext3.
>
> This sounds really odd. I can't imagine any form of filesystem
> corruption which could result in a directory full of data vanishing but
> e2fsck still being happy. Only an explicit user-land unlink ought to be
> able to do that. I've only seen this sort of thing reported once or
> twice in the past, but it has always ended up being traced to something
> happening in user space.
>
> Can you reproduce this with a current 2.4.21-pre kernel from Marcelo, by
> the way?
I can't reproduce it on demand, it is just something that seems to
happen. Could ext3 be cleaning things up automatically? I would suspect
that people don't really read the boot messages very carefully. I'll let
you know if it happens again.
Andy
--
-----------------------
Andy Vaught
TOLIS Engineering
andy@tolisgroup.com
_______________________________________________
Ext3-users@redhat.com
https://listman.redhat.com/mailman/listinfo/ext3-users
