ReiserFS and lost+found

[Roger Oberholtzer]

On Tue, 23 Apr 2002 13:21:41 -0400
Kurt Wall <kwall at kurtwerks.com> wrote:

> Scribbling feverishly on April 23, Roger Oberholtzer managed to emit:
> > On Tue, 23 Apr 2002 10:45:50 -0400
> > Kurt Wall <kwall at kurtwerks.com> wrote:
> > 
> > > directory. I have an all ReiserFS system at work and none of the
> > > filesystems have a lost+found directory. This is as of whatever patch
> > > set was necessary for kernel 2.4.5. I don't *think* Reiser uses a
> > > lost+found directory, but you could ask on the Reiser lists at
> > > namesys.com.
> > 
> > I will. Mine is for 2.4.13 (i.e., Caldera 3.1.1). The lost+found
> > directory is only used by the fsck utility, not the kernel. But, it must
> > exist and have lots of empty directory entries allocated. I will next
> > check the reiserfs fsck utility. There is no mention of lost+found in
> > that program's man pages. Next stop: source. I also see that the
> > mklost+found program only references ext2. But ext3 has a lost+found as
> > well. So I cannot trust that info as being complete. It is all up to the
> > fs-specific fsck utility.
> 
> I think the argument is that the journalled metadata can be replayed,
> obviating the neeed for lost+found. But, like you, I'm a tad
> uncomfortable with that notion. I checked with our filesystems guy
> here and he still doesn't totally trust ReiserFS. His recommendation
> is good enough for me. Obviously, *no* filesystem is perfect, but I
> tend to trust ext3 (ext2+journalling) or one of the more time-tested
> filesystems like XFS or OpenAFS.

We have had a bit of trouble with this Reiser setup. Of course, our
equipment is installed in vehicles that measure road surface roughness. So
the systems tend to bounce around a bit. But in 15 years, we have not had a
hard disk failure in any system. (Looking for wood. Knonk knock.) This could
just be bad luck. The Saudis are not keeping the roads up near the
border with Iraq in such great shape, but they still monitor their
condition. However, our roughness/speed measurements do not indicate that
the hard disk was overly bounced.

We will try ext3. Previously, we used UnixWare's Veriatis (bad spelling)
journaling FS, which I think is simply excellent. The Linux jfs are
a bit new to us. So, we are experimenting a bit as we go along.

-- 
+============================+===============================+
| Roger Oberholtzer          |   E-mail:        roger at opq.se |
| OPQ Systems AB             |      WWW:  http://www.opq.se/ |
| Erik Dahlbergsgatan 41-43  |    Phone: Int + 46 8   314223 |
| 115 32 Stockholm           |   Mobile: Int + 46 733 621657 |
| Sweden                     |      Fax: Int + 46 8   302602 |
+============================+===============================+