[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: A few quick questions regarding the journal.


On Thu, Jul 12, 2001 at 04:26:28PM +0200, Samuel Graenacher wrote:

> I've  just  recently noticed that the ext3 projects is being actiavely
> maintained  (again?). And looking at the new pages for 2.4.x it really
> made  me  interested.  But  I've got a few small questions I'd like to
> clarify before I start to use it.
> 1)  Tune2fs  and  mke2fs  talk  about a "default journal size" that is
> depending  on  the  partition  size but don't give any example.
> How big would the default journal be on a 20gb partition?

The appropriate journal size has nothing to do with the size of the
filesystem --- the journal gets recycled once it fills, so the best
journal size is really a function of the rate at which you use it up
(ie. what your workload is like) and the speed of your disk rather
than the filesystem size.  The default journal size should be fine,
although I'll probably do some testing later on to fine-tune the
journal sizes for different setups.

> 2)  Is it possible to remove the journal again? I'm aware that you can
> use  an  ext3  partition  with  an  ext2 kernel and have it completely
> ignore  the journal but I'm wondering if I could completely get rid of
> the journal again if I don't like ext3? (without reformatting ;))

Sure.  Just clear the "has_journal" flag on the filesystem, and e2fsck
will clear the journal out.  debugfs will let you clear the flag.

> (The  reason  for me asking is that the partition is usually kinda low
> on space and say a 32/64mb log would waste valuable space)

Actually, I usually use a journal nearer to 10mb on a lot of my

> >From  what  I've  read  the  the mailinglist archives tune2fs seems to
> create  a  filename  for  the  journal if used on a mounted partition.
> Would removing this file _cleanly_ remove the journal and make it ext2
> again? (doing that mounted as ext2 of course)

Sort of --- clearing the journal flag is much safer since it prevents
the kernel from thinking there's a journal and getting confused when
it doesn't find one.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]