Re: Silo and Raid1
On Wed, 2005-12-14 at 22:35 +0100, Turbo Fredriksson wrote:
> Quoting Simon Heywood <simon@triv.org.uk>:
>
> > On Wed, 14 Dec 2005 11:01:47 +0100, Turbo Fredriksson wrote:
> >> If you're going to use RAID on the disk, the first partition MUST start
> >> on block 1 (one!), not 0 (zero). Can't remember exactly why (usually I
> >> forget 'obvious reasons' :) but it have something to do with with ext2/3
> >> inode list, the boot block or with the RAID system needing block zero for
> >> something...
> >
> > The first few blocks of an ext2 or ext3 filesystem are unused, so
> > putting one at the start of the disk doesn't matter - the SILO code and
> > the disk label in block 0 of the disk are untouched.
> >
> > A partition that's part of an MD array will have data written to it from
> > its first block, so if it starts on block 0 of the disk then SILO and
> > the disk label be overwritten.
> >
> >> NOTE: This is true EVEN if the first partition isn't an MD!
> >
> > Why's that?
>
> Don't know. It's the facts.
>
> http://lists.debian.org/debian-sparc/2003/08/msg00123.html
> http://lists.suse.com/archive/suse-sparc/2002-Mar/0018.html
>
> I don't have time to prove my point. Believe me or not. Your call.
I fully agree. You also have to be careful with the last block of the
disk too. From memory the primary cause of these problems (when Raid or
LVM are concerned anyway) is the "whole disk" partition. The autoprobing
code finds this and in some circumstances decides to use the whole disk
rather than your partition with disasterous consequences.
Regards
Richard
--
Richard Mortimer <richm@oldelvet.org.uk>
Reply to: