So, I've yet to set up my mdadm raid 5 array, however, I will be very soon (In a couple days), and, I have everything down except for one issue that's rattling around in my head, the XFS documentation says it's better to match the stripe size & width up with your raid array to get 'significant' speed improvements, and, it shows you how to do it here, which, is all good and dandy until you want to add another hard-drive to your mdadm raid array, due to the fact that you're meant (from what I can see) to enter this data when you're creating the filesystem (mkfs.xfs), it leaves me wondering what happens when I upgrade my array to include more hard-drives. When I add a new hard-drive to the raid array, my stripe width will still be configured for my old, smaller, raid array, and, won't be in it's optimal state anymore. Is there a way to regenerate the superblock with the new stripe width (Stripe size, I'd presume, would stay the same)?

Thanks,
Automatic.

Why not just remount the filesystem and autodetect the new values as it suggests?

I must be really thick, may I ask where it says that? I see it says it'll autodetect, but, not on remount, only on creation.

EDIT:- Threw it onto VBox, made a small mdadm array using raid 5 & 3 8GB drives, aligned at 512KB. I made the xfs array, then, grew the mdadm array up to five drives, grew the xfs array, and, remounted it:-
http://i.imgur.com/0iCtHOV.png

/proc/mount contains:-
Which, after reading more about it here is in 512 byte sectors, so:-

sunit =1024 * 512bytes (Half a MB, 512KB)
swidth=2048 * 512bytes (One MB)

So, now, the sunit is correct, but, the swidth isn't (...I think?), however, if it's only a mount option I have to change I can deal with that. May I ask what exactly it changes? I was under the impression it was the whole foundations of the filesystem, not really something you can just 'slide around' in a couple of seconds remounting it.

EDIT:-
http://lxr.free-electrons.com/source/fs/xfs/xfs_super.c

That would lead me to believe that it's impossible to change, am I wrong? It clearly seems to only pick up on inode64/32, and, the write barrier option, along with a comment which says, well, you can read it for yourself.

So, does that mean it's impossible to change the value once the filesystem is created?