On 19/12/13 08:12, Wilson Jonathan wrote:

I realise if it is possible its not going to be the best solution, but a
temp stop gap...

What I am thinking of doing is while I save up to by my 4'th new disk
for a new raid 6 setup (4 drive raid 6, currently has 3 disks 1 missing)
is it possible to use 3 existing smaller disks, create a raid0 across
them and then add the resulting /dev/dmX as "a disk" where the currently
missing disk is marked as missing.

EG. /dev/md7  ... /dev/sdb1 /dev/sdc1 /dev/sdd1 missing

mdadm create /dev/md8 raid0 /dev/sde1 /dev/sdf1 /dev/sdg1

mdadm --add /dev/md7 /dev/md8

Yes, this is possible, MD doesn't care what the block device is.... MD 
uses block devices, and produces block devices...

If it is possible do I need to make sure the order in the conf file
first lists the /md8 then the /md7 or will mdadm see that a raid
contains a raid as a disk and then waits for it before assembling.

This might depend on what distribution you use /etc... IMHO, it should 
work without a problem, but you should try it to confirm that.

IMHO, this isn't such a terrible idea, it adds at least some level of 
protection, although if you can using a 4th smaller disk with RAID5 
would be better still. I'm also assuming performance is not a critical 
factor in all this...

BTW, can anyone comment whether this would work:
Using linear, then only the portion of the RAID0 with the failed disk 
will be failed in the RAID6, so 2/3rds will still be 4 disk RAID6 and 
1/3rd will be 3 disk RAID6?

Regards,
Adam

-- 
Adam Goryachev Website Managers www.websitemanagers.com.au