On Mon, 2003-10-20 at 20:45, James Sparenberg wrote:
> On Mon, 2003-10-20 at 10:42, Jack Coates wrote:
> > On Mon, 2003-10-20 at 10:23, Ricardo (Tru64 User) wrote:
> > > --- Jack Coates <[EMAIL PROTECTED]> wrote:
> > > > <soapbox>
> > > > if you've already backed it up, then go ahead and
> > > > try rebuild-tree; what
> > > > have you got to lose? I fully expect you'll have to
> > > > format and start
> > > > over, based on my experiences with ReiserFS, but
> > > > since you've got a back
> > > > up that's okay.
> > > 
> > > 
> > > Awww, it is a 900GB filesystem. Oh well, users, this
> > > FS is out of bounds for at least 24hrs!
> > > The corruption seemed to be on only one directory
> > > tree....
> > > 
> > yeah, so was the last one I saw melt down under ReiserFS. If it makes
> > you feel any better, rebuild-tree finished in about three hours the
> > first day, IIRC. It took about the same time the next day, maybe longer.
> > When it failed again the third day, it took about six hours to format an
> > ext3 filesystem and restore from tape, which was the last time anyone
> > had to worry about losing data on that system.
> 
> Interesting my experience is the exact opposite.  I've a number of boxes
> that I have to constantly "repair" due to ext3 (albeit they are redhat)
> The other boxes that do allow for reiser during install are all without
> hassle.  And these boxes get beat up hard. (The are all used for
> software testing.)  
> 
> James

This was RH 7.3 on a Dell 2650 with a PV220 array. It was the company's
CVS server with regular file-and-print duties as well. Maybe it was
something to do with the hardware, but the symptom was a kernel panic in
reiser.o under heavy load and the kernel errata from RH said it was a
bug in ReiserFS IIRC. The admin who had build the server was bummed,
having come from SuSE where Reiser is apparently more usable.

> 
> > 
> > > 
> > > [EMAIL PROTECTED]> ls -li *6345*
> > >   18541 -rw-r--r--    1 stuart user        1772 Oct 17
> > > 09:43 s_01_6345.subrun.ctl.2d
> > >   18542 -rw-r--r--    1 stuart user        1567 Oct 17
> > > 09:43 s_01_6345.subrun.ctl.3d
> > >   18543 -rw-r--r--    1 stuart user     3317856 Oct 17
> > > 09:43 p_02_6345.subrun.2d
> > >   18544 -rw-r--r--    1 stuart user    37602368 Oct 17
> > > 09:43 p_02_6345.subrun.3d
> > > 
> > > AND
> > > 
> > > [EMAIL PROTECTED]> ll -i p_02_6345.subrun.ctl.3d
> > >   18546 -rw-r--r--    1 stuart user        1567 Oct 17
> > > 09:38 p_02_6345.subrun.ctl.3d
> > > 
> > > So file with inode 18546 only shows up when listed
> > > explicitly. I could not rm/mv it either, claims no
> > > such file or directory, though i was able to "cp" it!!
> > > Strange!
> > > 
> > > 
> > >  You might want to consider using a
> > > > more stable FS next
> > > > time though.
> > > 
> > > Which is? (in your opinion, or proved ?)
> > > 
> > > _Thanks
> > > 
> > > Richard
> > > ...

interesting... never saw this message. Yeah for Sympa! I'll probably get
it on Friday.

I like ext3fs for desktop usage, but its performance is slow for
servers. On a server I'd look at XFS or JFS, depending on the work to be
done. IBM's developerworks site has some really good filesystem
comparison stuff. Hans Reiser's own site at Namesys also has some good
papers on general filesystem structure (for the record, I think ReiserFS
has a lot of potential, when its userspace utilities grow up).
-- 
Jack Coates
Monkeynoodle: A Scientific Venture...


Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com

Reply via email to