Thank you very much, Martin Bene and Bruno Prior, you have my gratitude!
All 6 raid1 devices work, and my system boots from the RAID devices,
autodetects them, and (so far!) works like a charm. The The
Software-RAID HOWTO v. 0.90.4 - Alpha, 17th of August 1999 is just a
little too terse for my limited understanding. I understand it more
clearly now. The main discoveries were:
1. the geometry, _including_ the start sector, are all available from
cfdisk. I changed the units to sectors, and, tada! there was the
starting sector for my disk section in lilo.conf. Even simpler:
you can get this by typing fdisk -lu. Thanks Harald Hansenfor your
example lilo.conf that you sent to the list.
2. I didn't realise that what the basic aim of using failed-disk is,
till I was doing it! It's really quite simple:
1. partitions with the precious data are marked as failed-disks
so that while they are technically part of the raid, in
practice they still exist as stand-alone disks that you can
copy your precious data from, to the raid
2. When you create the raid, it is stored on the other
partitions(s) (on the other disk). You format that and the
raid is empty (except for the lost_found directory).
3. You then copy the data from the "failed-disks" to the
formatted raid devices, just as described in the
Hard-Disk-Upgrade mini-HOWTO. In practice, the simplest way
(for me) was:
4. mount the root raid device under /mnt/root
5. cp -ax / /mnt/root
6. mount the other raid devices under /mnt/root
7. cp -ax /usr /mnt/root
8. cp -ax /var /mnt/root
9. etc.
3. The HOWTO does not show what a correct /proc/mdstat looks like
4. Alan Cox and Linus have decided to drop raid 0.90 from the kernels
for now. The latest kernel with raid 0.90 support is 504080 Aug 18
00:08 patch-2.2.12-final.bz2. The one that comes after that (even
more final?): 418815 Aug 20 07:24 patch-2.2.12-final3.bz2 is
smaller because (sniff sniff ;-( the raid code is lomotomised;
5. On Sun, 22 Aug 1999, Alan Cox wrote:
> > Nope. Enable auto-detection in the kernel (compiled in, or use
an
> > initrd) and you don't need any initscripts for raid.
> > I vote to leave it in.
> > (if this is a vote, otherwise, whatever. :)
> Andrea found a possible bug in the 0.90 raid code, and the more I
looked the
> less happy I was. Finally Linus thinks merging it is a bad idea.
So that
> discussion is settled
So has Andrea Arcangeli (?) told raid developers what the bug is,
and is someone working on it so we are freed from the slavery of
all having to individually hand-add patches to kernels? Besides, I
am using raid for reliability (I hate all the extra work when my
disk dies), so it would be nice to know of any bugs.
Nick Urbanik wrote:
> Dear folks,
>
> It is with fear and trepidation that I write this. I have two 14.4G
> disks, partitioned as shown below. I am compiling
> linux-2.2.12-final.gz ---apparently the last of Alan's kernels before
> he
> junked the raid patches ;-(
--
Nick Urbanik, Dept. of Electrical & Communications Engineering
Hong Kong Institute of Vocational Education (Tsing Yi)
email: [EMAIL PROTECTED], [EMAIL PROTECTED]
Tel: (852) 2436 8660, (825) 2436 8674 Fax: (852) 2436 8643
pgp ID: 7529555D fingerprint: 53 B6 6D 73 52 EE 1F EE EC F8 21 98 45 1C 23 7B
