From a cursory glance at beta3, rc.sysinit still includes the lines to
raidstart any arrays in /etc/raidtab. If you are starting your arrays
the conventional way, by marking the partitions type fd and getting the
kernel to autorecognize the arrays, they are already running by the time
rc.sysinit runs. This doesn't break anything that I could see, but
does produce plenty of warnings. Shouldn't the script check /proc/mdstat
to see whether the arrays are already running? Or did I miss the bit
where it did that? In an ideal world, the kernels should have RAID
support built in and scrap the RAID lines in rc.sysinit altogether.
And I had no joy getting the install to create RAID arrays. Had to do
the usual install to one of the RAID partitions, and then create the
RAID on the other partition with the first set as failed, copy the
filesystem across, modify fstab and raidhotadd the original partition to
the array. It's still the only practical way to get RAID running on
Mandrake. Diskdrake was happy to set things up for me, but when I
clicked Done, it came back with an error on mkraid. I would guess this
is because I was installing onto partitions that had already been used
for arrays previously, and that behind the scenes mkraid was requiring
--really-force to destroy the existing array. I am just guessing - I
don't know this is why it failed, there was no obvious clue on the
console. But if it is the reason, I guess the install should handle this
circumstance.
Cheers,
Bruno Prior
