I am in the process of trying to create a software raid using the "missing" disk option of mdadm. I'm following the instructions at:
I have two identical 18.4gig SCSI disks. I partitioned the first disk:
(I have /home on it's own separate drive.) The swap drive was formatted to Linux Swap and all the rest were formatted to ext3.
I then proceeded to do a base install of Debian Woody 3.0r1 on this disk with RAID compiled in the kernel (both raid1 and raid5) and I installed mdadm.
Using cfdisk, I duplicated the partitions of the first disk on the second. I set each of the partiton types on this drive to Linux Raid Autodetect (fd).
I then proceeded to create the RAID:
mdadm --create /dev/md1 --level=1 --raid-disks=2 missing /dev/sdb1
mdadm --create /dev/md2 --level=1 --raid-disks=2 missing /dev/sdb2
...and so forth for each of the partitions. I formatted each of the drives, mkfs.ext3 /dev/mdx for the regular partitions and mkswap /dev/sdb2 for the swap partiton.
This all proceeded fine and when done, cat /proc/mdstat showed the raid running fine (with the "missing/failed" devices of course)... I created mount points for each of the devices, i.e. /dev/md1 /mnt/md1 etc. I was even able to mount the devices and copy the file system from the "missing" devices to the mdx devices.
The problem enters in when I reboot the pc. I added the various devices and their mount points to fstab:
/dev/md1 /mnt/md1 ext3 defaults 0 0 and so forth for each of the md devices.
Then I reboot and the raid does not start... If I change the "pass" in fstab to "1" or "2", then I get an fschk error "unable to read superblock" error on each of the raid devices... and again, the raid does not start. (During the fs creation, I can see the superblocks being created...)
Any thoughts or ideas???
Thanks in advance for your help!