On Apr 29, 2002 11:37 -0500, Matt Stegman wrote:> I'm creating a large ext3 filesystem on an LVM volume, on top of a
RAID1
> for storing Norton Ghost disk images. Most of the files created will be
> between 1 and 10 GB; there may also be occasional small files also.
>
> I'm already going to reduce the number of inodes on the filesystem.
> Probably limit it to 102400 or so.
In that case you want to use something like "mke2fs -j -N 102400
<dev>"
> `mke2fs --help` reports a "-g blocks_per_group" option, but the
man page
> has no documentation on this. When I try to use it, any value I've
> specified returns "mke2fs: blocks per group count out of range"
The maximum size of block groups is currently fixed at 8*blocksize, so
32MB for 4kB blocksize. Changing the group size will have very little
effect on your performance given that with such a small number of inodes
in each group you are only skipping maybe 8 blocks (32kB) at a time.
Ted has proposed creating "meta block groups" which avoid this
specific
problem, but I'm not sure if it will have any noticable impact on
performance.
Cheers, Andreas
--
Andreas Dilger
http://www-mddsp.enel.ucalgary.ca/People/adilger/
http://sourceforge.net/projects/ext2resize/