Quick and to the point: the software RAID, despite booting properly and having everything appear to work, is useless. I disconnected the second disk in the RAID1 array and the OS refused to boot. I have no idea why. I think this has gone far beyond frustrating at this point. If I’m left with a RAID1 system that can’t even boot if a disk is missing then there’s no point having it the first place and I want my 250GB back.
My next step will be, having convinced the owner of the system to try out a proper hardware RAID card, to see if a hardware RAID1 will overcome these problems. I’ll have to look into it some more first of course and I’ll see if I can work out what is going wrong with the software implementation.
I also promise that not every post in the future will be about RAID!
UPDATE: A new plan has been formulated! I’m going to attempt separating out the /boot partition from mdadm’s clutches and then using rsync to make sure that both /boot partitions on each disk are kept in sync. Let’s see if that works.
I’d always thought the whole disk had to be in a RAID. If it does work, at least the boot partition won’t change much. Save for updating the kernel/grub.
For a hardware RAID, yes, you probably need to RAID the entire disk. But for mdadm, it works solely on partitions. You can even RAID partitions on the same disk!