> Suppose I have a SW RAID5 over 5 disks, 9.1 G each
> We are getting to a point where we need more space, but the box has no
> more space. So I am wondering how I might go about upgrading the disks to
> bigger capacity volumes. Suppose I insert bigger disks one by one and let
> the array reconstruct. Now when all of the disks have been replaced, will
> the md driver notice that it can expand itself? Or am I just kidding
> myself and I would need to recreate the filesystem in order to grow its
> size?
<pointlessly-theoretical-discussion>
IMHO, your up against a few pretty nasty problems. First as you
said, ext2 cannot be resized easily. So unless you run reiserfs or maybe
ext3 you're SOL right out of the gate.
If my understanding is correct, RAID offers a much worse outlook.
When a RAID volume is created, it checks the size of each drive and sets
the actual size of the RAID data on each drive to match the size of the
smallest drive in the set. This is so that all of the parity information
is properly distributed and no drive gets more than it's share. If you
remove a small drive and then add a larger drive and let is reconstruct
the RAID, the resulting size would remain the same.
You're best (but by this point purely academic) shot would be to
replace the old drives with larger ones one by one, where each of the new
drives had two partitions. The first partition would be the size of the
old raid partition, and the second would use the new space. Then you
could build another RAID-5 from the new partitions. At this point you
would have your old (n-times reconstructed) drive and a new drive, which
you could "Linear Append" together and then do a reiserfs online-resize.
On the other hand, RAID 5+Linear just might not work, infact it probably
doesn't.
</pointlessly-theoretical-discussion>
-Nic
