I have a fake raid, which I wanted to access using
mdadm /dev/md0 -A -c 128 -l stripe --verbose /dev/sda /dev/sdc
which should be right, as far as I understand the man page. But I get the message
mdadm: option -l not valid in assemble mode
leaving the offending option out leads to
mdadm: failed to create /dev/md0
and (despite verbose) no more information. I'm assuming that -A
requires some mdadm-specific header which is obviously missing. I probably need to use "build" instead of assemble, but from the description I'm really unsure whether this is a non-destructive operation. Is it? What should I exactly do?
UPDATE
I see I haven't made clear, that the array already exists as a fake-raid (I can't give the details about my mainboard now). It looks like doing nothing except for interleaving blocks, so I hoped it could be easily done using mdadm, too. Maybe I'm completely wrong, but all the info I've found was concerned with booting from fake-raid, what I don't really need. I'd be happy with a read access for now.
UPDATE 2
The solution from the answer nearly works. My fake-raid gets recognized by dmraid, but it refuses to start because of missing disks. I've forgotten to mention that my RAID10 is running without two disks i.e. it's just stripped. (I've already moved the important data to a safer place).
dmraid
. Could you update your post accordingly, please? – Oli Mar 04 '11 at 15:51dmraid
should pick up the controller and automatically make a mapping to a virtual device in/dev/mapper/
that you can mount. If you're trying to mount two fakeraid disks on another system, you're probably out of luck. fakeraid is painfully unmoveable. – Oli Mar 04 '11 at 17:05