Re: reiserfs on all my SRM-alpha drives: how I did it

[Michal Jaegermann <michal ellpspace math ualberta ca>]

On Sun, Oct 28, 2001 at 09:50:47AM -0500, Christopher C. Chimelis wrote:
> 
> Yes, I saw a few horror stories.  Luckily, this isn't really a production
> system, just a home machine that happens to have a partial mirror on
> it, so if it dies, I'm not worried.  But, the partition's contents only
> change once per day and it sure beats a 20 min. fsck :-)

I do not mean to imply that journaled file systems do not have their
uses.  But those who will understand them as a license to play rough
with file systems are replacement for backups may be for some nasty
surprises.

BTW - ext3 in its default, "ordered writes", mode practically prevents
accidents with a messed up content of files.  You may loose latest
updates on a journal replay but things should be sane.  In its "fast"
mode it behaves in this respect like reiserfs.  It also has "full
journal" mode which is expected to maintain a complete integrity of data
for a price of a drop in a performance and a bigger journal.

You would need updated e2fsprogs (the latest one from updates to
RH 7.1, i.e. e2fsprogs-1.23-1.7.1, should be ok and just recompile
if you have different libraries) and a kernel with a support for it
if you want to experiment with ext3.  A big attraction is that you
can convert ext2 filesystem to ext3, and back, without redoing the
whole filesystem from scratch.  Just use 'tune2fs'.  Also booting
an older kernel without ext3 support will simply mount even converted
filesystem as ext2 type and will not use journal (but I would think
that in an event of an unclean shutdown when you had it mounted as
ext2 replaying journal on the next boot would be not a good idea :-).

  Michal