Debian/raid1/mdadm/devices not created at boot time

Debian/raid1/mdadm/devices not created at boot time

[Samuel Tardieu]

Hi.

I have migrated my Debian -unstable system to raid1 handled by madm as
directed in http://xtronics.com/reference/SATA-RAID-debian-for%202.6.html.
I use Linux 2.6.9, / is on md2, /boot on md0 and swap on md1. I use
mdadm 1.7.0 (the one shipped with Debian).

At boot time, the kernel correctly boots from md2 but init fails to
mount md0 as /boot and setup the swap space on md1 as those devices do
not exist.

If I create them by hand with "cd /dev && ./MAKEDEV md" when I get a
root shell for maintenance and start them using "mdadm -A -s", I can
continue the boot process manually.

The partition types are set to 0xfd, /proc/mdstat reports:
Personalities : [raid1] 
md0 : active raid1 sda1[0] sdb1[1]
      192640 blocks [2/2] [UU]
      
md1 : active raid1 sda5[0] sdb5[1]
      1951744 blocks [2/2] [UU]
      
md2 : active raid1 sda6[0] sdb6[1]
      37110016 blocks [2/2] [UU]
      
unused devices: <none>

As far as I can tell, the "mdadm-raid" script which does "mdadm -A -s"
is present as /etc/rcS.d/S25mdadm-raid which is a right place as it
will be run before mounting other file systems. However, the absence
of /dev/md0 and /dev/md1 prevent the raid arrays from being started.

What can I do to ensure that /dev/md0 and /dev/md1 exist when the
script is run?

  Sam
-- 
Samuel Tardieu -- sam@rfc1149.net -- http://www.rfc1149.net/sam

Re: Debian/raid1/mdadm/devices not created at boot time

[Samuel Tardieu]

> I just overflew the link you gave. There the partition type "fd"
> approach is used, which needs either compiled-in raid functionality or
> raid as a module in the initial ramdisk (which is both at least fo Woody
> not the standard case; I built a bootable cd with a kernel with
> compiled-in raid support so that I could install directly onto a raid).

I have raid as a module in the initrd.
 
> If you want the disks to be started by the mdadm call in that script,
> then the devices shouldn´t be assembled (or even tried to) by
> autodetection before. Try changing the type back to 83/82 or whatever fs
> you use.

It won't change anything (I had 83 before I noticed I forgot this
part), but I understand what you're saying.

> What exactly does lead you to the thought that /dev/md0 and /dev/md1 do
> not exist? Are the actual entries in /dev issing? If so, why are they
> missing again after you made them wit MAKEDEV?

When I get the maintenance root prompt, I see that /dev/md0 and
/dev/md1 are missing from /dev. When I reboot, they have disappeared
again.

Although nothing appears in mnttab, I suspect my system may be using devfs:

willow# df -k /dev
Filesystem           1K-blocks      Used Available Use% Mounted on
none                      5120      2944      2176  58% /dev

At boot time, no entries are created in this devfs for /dev/md0 and
/dev/md1. How can I create them? I thought loading the md module from
initrd would be enough but it looks like it isn't.

  Sam
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Debian/raid1/mdadm/devices not created at boot time

[Norman Schmidt]

Samuel Tardieu schrieb:
> Hi.
> 
> I have migrated my Debian -unstable system to raid1 handled by madm as
> directed in http://xtronics.com/reference/SATA-RAID-debian-for%202.6.html.
> I use Linux 2.6.9, / is on md2, /boot on md0 and swap on md1. I use
> mdadm 1.7.0 (the one shipped with Debian).
> 
> At boot time, the kernel correctly boots from md2 but init fails to
> mount md0 as /boot and setup the swap space on md1 as those devices do
> not exist.
> 
> If I create them by hand with "cd /dev && ./MAKEDEV md" when I get a
> root shell for maintenance and start them using "mdadm -A -s", I can
> continue the boot process manually.
> 
> The partition types are set to 0xfd, /proc/mdstat reports:
> Personalities : [raid1] 
> md0 : active raid1 sda1[0] sdb1[1]
>       192640 blocks [2/2] [UU]
>       
> md1 : active raid1 sda5[0] sdb5[1]
>       1951744 blocks [2/2] [UU]
>       
> md2 : active raid1 sda6[0] sdb6[1]
>       37110016 blocks [2/2] [UU]
>       
> unused devices: <none>
> 
> As far as I can tell, the "mdadm-raid" script which does "mdadm -A -s"
> is present as /etc/rcS.d/S25mdadm-raid which is a right place as it
> will be run before mounting other file systems. However, the absence
> of /dev/md0 and /dev/md1 prevent the raid arrays from being started.
> 
> What can I do to ensure that /dev/md0 and /dev/md1 exist when the
> script is run?
> 
>   Sam

Hi!

I just overflew the link you gave. There the partition type "fd"
approach is used, which needs either compiled-in raid functionality or
raid as a module in the initial ramdisk (which is both at least fo Woody
not the standard case; I built a bootable cd with a kernel with
compiled-in raid support so that I could install directly onto a raid).

If you want the disks to be started by the mdadm call in that script,
then the devices shouldn´t be assembled (or even tried to) by
autodetection before. Try changing the type back to 83/82 or whatever fs
you use.

What exactly does lead you to the thought that /dev/md0 and /dev/md1 do
not exist? Are the actual entries in /dev issing? If so, why are they
missing again after you made them wit MAKEDEV?

Norman.

-- 
Norman Schmidt          Institut fuer Physikal. u. Theoret. Chemie
Dipl.-Chem. Univ.       Friedrich-Alexander-Universitaet
schmidt@naa.net         Erlangen-Nuernberg
                         IT-Systembetreuer Physikalische Chemie

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html