Subject: Re: Paramaters for large drives?
To: Rick Byers <rb-netbsd@BigScaryChildren.net>
From: Brian A. Seklecki <lavalamp@spiritual-machines.org>
List: netbsd-help
Date: 08/13/2002 23:47:04
On Mon, 12 Aug 2002 @ 6:11pm (-0400), Rick Byers wrote:
RB> I think I've got it working now, but I want to make sure I'm not totally
RB> out to lunch here. Here's what I did:
RB> - changed the geometry to increase the number of cylinders and decrease
RB> the number of heads: 35556 cyl, 64 hd, 63 sec
RB> The boot code is accessed by block number, so it shouldn't matter
RB> that this doesn't match the BIOS geometry, right?
RB> - use 8192/1024 bs/fs on all reasonably sized (<=6GB) slices
...there were a number of threads a few months ago about large drives (160
gig was in question, you know the maxtor's). Let me see if I can dig them
up.
-lava
RB> - for my big slice (62GB):
RB> - use 16384/2048 bs/fs
RB> - use the max cpg (99) - 359 groups (instead of the default ~2000)
RB> - use -i 16384 so I get a sane number of inodes (4M)
RB>
RB> Sound good?
RB>
RB> Thanks,
RB> Rick
RB>
RB> On Mon, 12 Aug 2002, Rick Byers wrote:
RB>
RB> > Hi,
RB> > I've got a 70GB raid set that I want to install NetBSD-1.5.3/i386 onto.
RB> > The logical disk has 8924 cyls, 255 heads, and 63 sectors. I want to make
RB> > a "small" root partition (512 Meg). I used the default of 8Kb block
RB> > size and 1Kb fragments (since the root partition will probably have a lot
RB> > of little files on it, I'd rather not use a huge fragment size).
RB> >
RB> > Newfs gives an error of "With 16065 sectors per cylinder, minimum
RB> > cylinders per group is 64. This requires a block size of 16386 and
RB> > fragment size of 4096." I don't really want to use such a large frag
RB> > size. But if I try that, I get a warning about the rotational map not
RB> > fitting in the super block and filesystem performance being degraded.
RB> >
RB> > What sort of parameters should I use for such a large drive? I'm not too
RB> > concerned about wasting some space (I've got way more than enough), but I
RB> > don't want to end up with horrible performance.
RB> >
RB> > Thanks,
RB> > Rick
RB> >
RB> >
RB>
RB>
later - | _BonaFide[] = { coder, author, | /~\ The ASCII Ribbon
Brian A. Seklecki | problem solver, scholar, BOFH, | \ / Campaign Against
| vegetarian, runner, NetBSD | X Exchange, Outlook
| advocate, spiritual machine }; | / \ & HTML Email
PGP: 0111 4618 0111 4618 4AB2 24EB C853 6611 75C4 A91E A7A7 4973
"GNU/Linux: About as stable as the elements at the bottom of the periodic table"