Strange raid device numbering

Strange raid device numbering

[J.A. Magallón]

Hi...

I'm running kernel 2.6.36-rc6.
I have just built a simple raid0 array with two disks, and I followed what I
had always done:

raid=/dev/md0
ndisk=2
disks="/dev/sda1 /dev/sdb1"
block=4
chunk=256

stride=$((chunk/block))
stripe=$((stride*ndisk))

for i in $disks
do
    mdadm --zero-superblock --force $i
done
mdadm -C -f -v $raid --level=0 --chunk=$chunk -n $ndisk $disks
sleep 5
mdadm -S $raid
mdadm --assemble $raid $disks
mkfs.ext4 -b $((block*1024)) -E stride=$stride,stripe-width=$stripe $raid

Strangely, after I reboot, the system insists in numbering it as
md127, instead of md0.

Why ?
How can I force it to be md0 ? (its half aesthetics, half curiosity...)

TIA

-- 
J.A. Magallon <jamagallon()ono!com>     \               Software is like sex:
                                         \         It's better when it's free

Re: Strange raid device numbering

[Phil Turmel]

Hi J.A.,

On 10/07/2010 07:43 PM, J.A. Magallón wrote:
> Hi...
> 
> I'm running kernel 2.6.36-rc6.
> I have just built a simple raid0 array with two disks, and I followed what I
> had always done:

[snip /]

> Strangely, after I reboot, the system insists in numbering it as
> md127, instead of md0.
> 
> Why ?
> How can I force it to be md0 ? (its half aesthetics, half curiosity...)

I'm gonna guess that you have some combination of an initramfs, and/or you
don't have an /etc/mdadm.conf, and/or you haven't updated your initramfs with
the latest copy of it.

Raid auto-assembly in the kernel uses the minor numbers in the superblock (if
it can).  Assembly by mdadm without instructions from mdadm.conf uses minor
numbers from 127 and down.  (This may have changed in mdadm in the last few
months, but I recently switched from kernel auto-assembly to initramfs
assembly, so I can't say for sure.)

I've chosen to *not* use mdadm.conf, and find all of my devices with "LABEL="
syntax.

HTH,

Phil