> 1. I made a RAID-0 array (raidtab & mkraid /dev/md0) under Mandrake
> 7.0-2 using a fresh 2.2.14 kernel with IDE & RAID patches.
NB: RAID0 and linear I would not expect to do anything.
RAID1, RAID4 and RAID5 are likely to scribble over some of the disks
[ e.g. RAID1 will copy one disk to the other disk(s) ]
> 2. Moved the drives so they were in different positions and booted.
If you had them marked as type 0xfd, had PSBs enabled and had auto-detection
enabled in the kernel, it should worked.
> 5. Without doing a mkfs or even a fsck I mounted /dev/md0 and was
> able to access all the data again.
RAID1 should not care as long as things were closed down correctly -- all
disks should have the same (non-meta) data, so whichever one is copied to
whichever other one, it should not matter.
Otherwise, if you kept the disks in the same order, then this should work.
Otherwise, RAID0 and linear should be OK when put in the correct order
I'm not sure about how the XOR stuff will get on if you shuffle disks in a
RAID4 or RAID5 setup ... I'd not want to risk it myself !
> From this it seems that mkraid just modifies the RAID superblock (if
> there is such a beast) and not the superblock of the filesystem.
Indeed -- it is proving a virtual block device, so has no concept of a
filesystem so can't trash the superblock.
But it may well write lots of disk blocks for anything other than RAID0 or
linear
