Journaling file systems revisited

[Glenn Burkhardt]

> So, in general, I would assume that a journalling file system does not need
> a periodic equivalent to the fsck. 

Correct.  That's the point of the journaling.

> ReiserFS works well for some, but a Google search on "reiserfs
> corruption" (an AND search) produces 3000 hits. "ext3 corruption" gives
> you 400 hits, and "xfs corruption" just 123. I wouldn't call this a
> scientific measure, of course :-)

> After ReiserFS chewed up my /home partition last year, I found online
> it's happened to lots of folks. It's just my personal opinion, but I
> would look between XFS, and ext3 (perhaps JFS... I haven't looked at JFS
> yet)

The search results probably only reflect the number of users of the various 
filesystems, or perhaps in the case of ReiserFS, early development, since 
that filesystem was built from scratch, and to do more than journaling.

Within the last year there were reports of poor performance of the ReiserFS 
when the same computer was used as an NFS server.

The XFS system was developed by Silicon Graphics, and used with their IRIX 
system.  So it was developed by paid professionals, with commercial bug 
tracking, and used commercially (does this make it better?).

JFS was donated by IBM recently (June 2001) and is now at version v1.0.20 as 
of 06/21/2002.  So it's new, but presumably IBM wanted it so that Linux on 
its servers ran with a vetted (by them) journaling filesystem.

The ReiserFS was the earliest (I think!) journaling filesystem for Linux, and 
was supported early on by SUSE.  But probably for any distribution released 
now, any of the filesystems ReiserFS, ext3, XFS, or JFS will give what one 
wants most - filesystems that are almost immune to power outages.

That's what I'm most interested in, at least, since my company uses Linux 
systems in our products, and I hate trying to tell users over the phone to 
type 'fsck /dev/hda1' when their systems won't boot.  Especially if their 
native language is not English (or systems speak, for that matter).  Computer 
systems should always boot, unless the hardware is faulty.

P.S.  Some of our older systems have only ext2, and have gotten into a funky 
state that fsck doesn't fix.  No files can be created (file system full), but 
existing files can be read.  'df' reports plenty of free space, and I can't 
find unusually large files in /tmp, /var/tmp, or /var/spool.  I've tried to 
force an 'fsck' with an abrupt powerdown, and the user reports the file 
system being checked (I can't look myself, the system is in Italy).  So no 
temporary files can be created, and some things don't work.  What can I do, 
short of sending a new hard drive, or re-building the file system 
(re-installing Linux)?