also sprach Tony Godshall <togo@of.net> [2009.05.11.2058 +0200]:
> They are *not*. That would seem to be the nub of the problem. No
> base partitions, thus no RAID. Good, thanks, that's progress. Might
> be good for md to report that that a little better (see screenshot).
> Oh, duh, I moved /etc/mdadm/mdadm.conf out of the way. So it doesn't
> even know I have a RAID until it autodetects it, right? Oh, wait, in
> this initramfs environment there is an /etc/mdadm/mdadm.conf, and it
> shows UUID and the 4-way RAID1. What, so the /etc/mdadm/mdadm.conf
> inside the initrd is not from the kernel package but was built
> locally? Didn't realize that.
It's copied from the main system when the initramfs is created,
unless it's faulty or doesn't represent the system or hasn't been
checked since the 2.5.3 upgrade.
In the initramfs, if there is a config file, it's used. If there's
not a config file, mdadm scans all partitions and assembles them all
(like /sbin/mdadm-startall).
Thus, if you remove it in the initramfs, you basically tell the
system that you want it to scan, rather than search for the specific
UUID in the file.
> Perhaps if I could figure out how to grab the dmesg output from
> initramfs environment, I could then compare the amd64 kernel
> output to the 686-bigmem kernel output... :-/
Append 'break=bottom debug' to the kernel line and at the shell do
something like
mount -o remount,rw /root
cp /dev/.initramfs/initramfs.debug /root/root/initramfs-debug.foo
mount -o remount,ro /root
and then find the file in /root once the system booted.
--
.''`. martin f. krafft <madduck@d.o> Related projects:
: :' : proud Debian developer http://debiansystem.info
`. `'` http://people.debian.org/~madduck http://vcs-pkg.org
`- Debian - when you have better things to do than fixing systems
review of a chemistry paper:
"paper should be greatly reduced or completely oxidized."
-- frank vastola
Attachment:
digital_signature_gpg.asc
Description: Digital signature (see http://martin-krafft.net/gpg/)