Hi Gordon,
I tries every trick under the sun, cpio, cp -ax, tar, rsync...
As soon as I updated to testing all files copied on the same drive same
partition, same raidtab etc. etc.
All the files copied without failure, including symlinks, which noteably
failed with the previous 15 frustrating attempts.
Hmmm.... Not sure where the problems came from, did you format your
/mdx using mke2fs?
I've built several boxes to boot off raid1 using Debian woody. Eg:
drogon @ agate: df -h
Filesystem Size Used Avail Use% Mounted on
/dev/md0 235M 18M 205M 8% /
/dev/md1 1.9G 905M 920M 50% /usr
/dev/md2 35G 12G 21G 34% /var
drogon @ agate: uname -a
Linux agate 2.4.20 #1 Wed Jan 29 12:07:57 GMT 2003 i686
unknown drogon @ agate: cat /etc/issue Debian GNU/\s 3.0 \n \l
Theres a good howto on it all at
http://www.tldp.org/HOWTO/Software-RAID-HOWTO-4.html#ss4.11
The trick to doing the copy is to use cpio (or something else that
preserves device nodes rather than try to crete files >out of them)
Gordon
On Fri, 21 Feb 2003, Lewis Shobbrook wrote:
Hi Gordon,
I tries every trick under the sun, cpio, cp -ax, tar, rsync...
As soon as I updated to testing all files copied on the same drive same
partition, same raidtab etc. etc.
All the files copied without failure, including symlinks, which noteably
failed with the previous 15 frustrating attempts.
Hmmm.... Not sure where the problems came from, did you format your
/mdx using mke2fs?
Yes. When I get time I want to look at XFS, but last time I looked (about
6 weeks ago) they didn't have patches for 2.4.20, and I needed a
particular PCI-IDE card that only 2.4.19 and upwards seemed to
support )-:
quoted
drogon @ agate: df -h
Filesystem Size Used Avail Use% Mounted on
/dev/md0 235M 18M 205M 8% /
/dev/md1 1.9G 905M 920M 50% /usr
/dev/md2 35G 12G 21G 34% /var
This particular box has 2 disks all on raid1 mirrors. I started by
partitioning and installing onto 4 primary partitions, hda1, hda3 and hda4
(hda2 is swap, but only later on). Once I had the basic system installed
on the primary disk, and built a custom kernel and tested it, I used fdisk
to create an identical partition table on hdc, then made hda2 and hdc2 an
md device, mkfs'd it, mounted it under /mnt and used cpio to copy / into
it. Then changed the /mnt/etc/lilo.conf, ran lilo -r /mnt, changed
/mnt/etc/fstab (/ is now under /dev/md0!) and rebooted. Once that was OK,
I made an md device out of /dev/hda1 and /dev/hdc1 and basically reversed
the procedure to move root back onto the first partition on the disks.
then I used the now spare partition to temporarily copy /usr and /var
while turning them into md devices. It takes a long time and I wish I
could do it at install time like Red Hat does, but maybe Debian 4 might
have that in it!
I've built several like this, some with raid1 for / and raid5 for the
other partitions, depending on the number of disks installed and the
customer requirements.
I don't use a separate /boot filesystem - never quite understood that, but
my unix dealings before Linux was SunOs & Solaris...
I use the raidtools as suplied with woody - I am not using madm at all
(old habits die hard and I can hand-edit an /etc/raditab without thinking
about it)
Gordon