What is the best way to increase the size of a "full" Xserve raid?

Question

Hi,
our Xserve raid consists of 2 x 7 x 350GB disks.
These are arranged in 2 volumes (one per controller), RAID 0+1, 3 disks striped, then mirrored, and such are using 6 of the 7 disks per controller. 1 disk per controller is configured as a hot spare.
During time both volumes have filled up and now there only are 100GB left on every volume and we need more space
Is there a way to hot replace 350GB disks by 750GB disks?
Or is the only way to copy data off, replace disks, and restore data?
I would guess, that I can replace one disk with a 750GB disk, but only 250GB of it will be used. If I then would wait for reconstruction of that new part, and then replace its counterpart? Would only the first 350GB be reconstructed and the 400GB be lost, or would the raid recognize the free space and allow to add it to the stripes and mirrors?
Cheers,
Martin

Answer 1

A disks in the array need to be the same size, so your guess is correct - if you replace a disk with a larger disk, the additional space will be lost.
As a result there's no trivial way of doing this.
Copying off the data, reformatting and restoring the data is the best way to go.
If that's not possible, you have one small chance thanks to how you've currently configured the arrays as 0+1.
You could break the mirror on controller 1- that would give you two logical arrays of three disks each. You could then take the data off the 0+1 array on the second controller and put it on the second half of the first controller, leaving the second controller's volumes spare.
Then destroy both volumes on the second controller and rebuild it as a RAID 5 volume. If you want to retain a hot-spare, this will equate to a 6-disk RAID 5, or about 1.8TB
Now you move the data from the two RAID 0 arrays on controller 1 onto the RAID 5 array on controller 2 and repeat the process to rebuild the 6 disks into a single RAID 5 volume. You can then redistribute the files as you like.
Of course, during this process you'll be running with lower (read: no) redundancy, so you had better hope a disk doesn't fail in the process.
A 6 disk RAID 5 is going to offer about the same performance and redundancy as the two mirrored 3-disk RAID 0 arrays (more performance if those RAID 0 arrays are full), with the advantage that RAID 5 isn't so wasteful of disk space (you lose have your total space to the mirror as opposed to one disks-worth of data to parity in RAID 5).
Either way, back the data up first.