On Thu, Feb 15, 2007 at 07:27:22PM -0600, Mike McCarty wrote:
> [...]
> >maybe you should read about LVM [1]. It is not about file systems, but
> >it can help you :)
>
> I'd rather deal with a case of the Clap.
>
> LVM is worse than useless for most installations. It makes
> the entire file system dependent on every drive in the Logical
> Volume working. If any drive fails, then the entire FS becomes
> corrupt. As you may know, as the number of devices goes up,
> the MTBF goes down drastically, and the probability of failure
> goes up dramatically. If one has a largish RAID, then LVM makes
> sense, but without RAID or some other error correcting ability,
> LVM makes the likelihood of a file system failure increase, and
> makes the likelihood of recovery from it decrease, since the
> normal recovery tools won't work.
> [...]
Actually this is correct only if one chooses to use LVM to stripe the
physical volumes. But nobody said it must be done that way. I use LVM
too and have two volume groups, one for each HD which is completely
managed by LVM. So each VG just contains one physical volume, So if
one HD fails, well shit happens, but that does not destroy the data on
my other VG.
And LVM provides more flexibility for partitioning. If I need some
more space on one partition and have another unecessarily big one, I
can shrink the filesystem on the bigger on, then shrink the logical
volume as well and afterwards hand over the free extents to the
partition, where they are needed and grow the filesystem appropriatly.
Plus I don't need to shift the partions, to get the wanted space
before or after the partition I want ot grow. So that is IMHO a big
gain in flexibility. Of this adds a little amount fragmentation, but
after some shifting extents here and there I have not recognized any
significant performance decrease.
So this is where my suggestion for a filesystem comes into play. I
used XFS in the beginning of my experiments with LVM but am migrating
to ext3 now, since XFS can only be grown but not shrunk. But growing
*and* shrinking are both natively supported features of ext3. Also it
seems to be slightly faster than XFS on my setup. I recently migrated
my /home (about 16GB) partition from XFS to ext3, using some spare
diskspace with an intermediate ext3 partition. Copying from the XFS to
the ext3 partition took about 14 minutes, whilst copying from the
intermediate ext3 to the final ext3 /home took 12 minutes.
Regards
--
Marcus Blumhagen
"Any intelligent fool can make things bigger, more complex, and more
violent. It takes a touch of genius -- and a lot of courage -- to move
in the opposite direction."
-- Albert Einstein
Attachment:
signature.asc
Description: Digital signature