This One Time, at Band Camp, Florian Philipp <lists@f_philipp.fastmail.net> said, On Sat, Feb 16, 2008 at 01:50:04AM +0100:
> On Sat, 2008-02-16 at 00:32 +0100, Wael Nasreddine wrote:
> > > To your filesystem scheme: Why do you use xfs for usr? AFAIK XFS is good
> > > at write speed but not worth the trouble when reading data and data in
> > > usr is usually written once, updated every few months and read many
> > > times a week (on rebooting Desktop PCs maybe once a day). I'd use
> > > reiserfs3.6, maybe even without notail to make it more space efficient.
> > I don't use XFS, curently I only have / and /home and I want to split
> > it to more smaller partitions, I'm on LVM so it's easy, anyway I'm
> > going with ReiserFS for /usr /var, would you please suggest
> > mkfs.reiserfs options as I have nerver used ReiserFS-3 before (yep 5
> > years using linux and I've always used ext3...) also You didn't mention
> > /var, would you say ReiserFS-3 is a good choice as well?
> I don't think there's alot to do when creating a reiserfs. You could
> change the number of blocks for the journal. A bigger journal allows
> larger transactions which speed up write actions but might waste space.
> If you've got a second hard drive you could use an external journal but
> I've never done any benchmarking on that issue although I use it on my
> personal wannabe server (a raid1 and a single disk for the journal and
> unimportant data).
> I didn't comment on /var because I don't know how you use it. I suspect
> it to hold alot of temporal data like lock files, spools and so on. So
> there's a lot of creating and removing files going on, possibly in
> parallel. XFS is good in parallel and in creating files but terrible in
> removing files. Reiserfs with notail seems a good choice if you ask me
> (what you did
> > > I'd also use ext2 on /usr/portage. These data don't need journaling.
> > > Everything's got an MD5-sum to make sure it's unchanged after a crash
> > > and you can easily resync. I found ext2 with 2k blocks to be faster than
> > > reiserfs3.6, even on read-performance.
> > I've already made the partition as suggested in  I used this
> > command:
> > $ mke2fs -b 1024 -N 200000 -m 0 -O dir_index
> > I guess 1K block size would be faster??
> I'm not sure. 2K blocks might reduce fragmentation.
> If you look at the output of
> find /usr/portage/ -type f | xargs du -h --apparent-size
> you'll see that there are quiet a few files larger than 1K but most are
> smaller and might stay that small. So yes, I think 1K is a good choice
> but you won't loose much with 2K, maybe you even gain some speed.
> > > If I were you, I'd also use separate volumes for /tmp and /var/tmp
> > > (without ccache) with xfs.
> > What did you mean by 'without ccache'? I have ccache and I use it...
> I meant that you should keep ccache on a separate partition. I just
> think: Less stuff in the FS, less work on allocation and lookup, more
> speed. And there's a lot of stuff in 2GB ccache.
> By the way: I don't think /var/tmp is a good place for ccache (not
> technically, just for the sake of layout). I've moved it to /var/db
> since it's not really a bunch of temporary data but more like a changing
> > > /home could use data=journal. Those data are precious and if I remember
> > > correctly, this setting even brings an obscure (i.e. undocumented) speed
> > > improvement with many parallel disk accesses, for example in a
> > > multi-user environment.
> > it's done, thanks, BTW what's your home partition FS? your choice is
> > ext3 or reiserFS??
> I use reiserfs3.6 without notail but that doesn't mean that it would be
> a good choice for you. I'm on laptop and disk space efficiency is a big
> topic for me so I use tail-packing wherever suitable. And yes, I am a
> fan of ReiserFS-3.6. I think it's the best multipurpose FS. You can
> easily adapt it for high performance or high disk space efficiency. If
> its journaling would be as good as Ext3's data=journal I'd use it
> everywhere except for small partitions (ext2) and big files (ext3 and
> > One last thing, since I'm on LVM resizing the partition is a must
> > feature, in ext3 I use resize2fs which works quite nicely, is
> > resize_reiserfs as reliable as resize2fs is??
> Yes, it's just as good and the sky's the limit for resizing
> Oh, by the way: If you choose to use XFS somewhere, keep in mind that
> you can't shrink and XFS-FS. Neither online nor offline.
> One last thing: It's a bit old but I think it's still interesting,
> especially for XFS-users:
Thank you for your detailed answer it helped a lot, I just finished
resizing/migrating all partitions, Though I still have the Storage
partition, which is for my Mp3z and is almost 70Gb, with ext3, I'll
see later if I do migrate to ReiserFS or not but the rest is done,
please take a look at the file attached... and if you have any more
suggestions please do tell me.
Thanks a lot guys....
PGP: 1024D/C8DD18A2 06F6 1622 4BC8 4CEB D724 DE12 5565 3945 C8DD 18A2
.: An infinite number of monkeys typing into GNU emacs,
would never make a good program. (L. Torvalds 1995) :.
/dev/system/gentoo-root / ext3 defaults 0 1
/dev/whd120.1 /common-boot ext2 defaults 0 0
/dev/system/home /home reiserfs defaults,user_xattr 0 0
/dev/system/gentoo-overlays /home/overlays ext2 defaults 0 0
/dev/system/gentoo-usr /usr reiserfs defaults 0 0
/dev/system/gentoo-var /var reiserfs defaults,notail 0 0
/dev/system/tmp /tmp xfs defaults 0 0
/dev/system/var-tmp /var/tmp xfs defaults 0 0
/dev/system/storage /mnt/storage ext3 defaults,user_xattr 0 0
/dev/system/swap none swap sw 0 0
/dev/cdrom /mnt/cdrom iso9660 noauto,ro,user 0 0
proc /proc proc defaults 0 0
shm /dev/shm tmpfs nodev,nosuid,noexec 0 0
# Sizes: (120Gb Hdd)
I use gentoo/paludis building on another partition... the gentoo-overlays partition is
the one for /usr/portage but actually I use paludis and so I have all overlays (gentoo and
whatever layman gets) in /home/overlays...