Re: [ext2resize] Q: ext2resize working for ext3?

[Andreas D. <ad...@tu...>]

On Nov 15, 2001  07:27 +0100, Ph. Marek wrote:
> I'm very impressed of the possibility of online-resizing ext2 (using the
> kernel-patch). But as of 2.4.14pre4 ext3 has been included into the
> -linus-kernel and ext2 being the same as ext3 (disk-layout meant) does the
> resizing (kernel-patch + utilities) work for ext3 too? (preferably *online*)

The offline resizing works fine, as you suspected.  (Note that you don't need
the kernel patch for offline resizing of ext2 or ext3).  However, the online
resizing does not work for ext3 yet.  Almost a year ago, I worked on doing
online ext3 resizing (for 2.2 kernels), and it was mostly done, but I have
never ported it to 2.4 ext3 yet.  It needs some small changes to the format
of the reserved node, so ext2prepare needs to be changed also.

More and more people have started bugging me about this, so I will probably
start working on it again.

> I'd expect that ext2resize would have to resize the journal too - which
> it's likely to have problems with if it's on another device instead of just
> a file in the filesystem, but it should be doable - especially if you're
> using lvm (which is the point where online-resizing makes sense, after all).

Well, I won't try to resize the journal at all.  It would be possible, but
it is also possible to just delete the journal and create a new one.  Since
it is not _required_ to have a larger journal for a larger filesystem, it
is just more complexity that adds the possibility of a problem.  I don't
think that it would be very easy to do online journal resizing, and once
you unmount the fs, you can just use tune2fs to delete and re-create one.

Also this would possibly fragment the journal a lot (normally the journal
is at the start of the fs, but if the fs is full, the only free space
will be at the end, so you will seek the whole disk while writing to the
journal, ugh).  You are far better off to delete the journal, and create
a new one, which will probably be contiguous in the new group.

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/