Leslie Rhorer wrote:
I have a few questions. Some RAID implementations will simply
refuse to create or grow an array if all the targets are not precisely
the
same size. Clearly this is not the case for mdadm. Not all drives of a
given "size" are actually precisely the same size, however, and I am
using
unpartitioned drives for my RAID systems. What happens if I add a drive
whose apparent physical size is a bit smaller than the device size used
to
create the array?
For RAID 4/5/6, I think it'll be refused.
Do you know if the refusal would include an error message clearly
indicating why the growth is refused?
You have to shrink the
filesystem, and LVM if you use it, then the array, so the used size is
no bigger than the new drive - as you've noted, md doesn't mind if it
doesn't use all the available space on its constituent devices. If it's
a small reduction, as I imagine it would be, and your filesystem
supports shrinking, it won't take long to do the the shrinks. Then
adding the new drive will be painless. If your filesystem won't shrink -
and some (many?) won't - I suspect you're scuppered.
I'm no longer using LVM on any of the servers, and I've converted to
XFS on RAID 5 and RAID 6 arrays. At this time XFS does not support
shrinking. I've seen some chatter on the web about 3rd party utilities
which might make it possible.
For growing an array, this would be a bit of a pain, but probably
not a show stopper. Even for a failed drive I could probably just send the
new drive back and purchase a different model whose real size is as large or
larger than the extant drives. The problem is, waiting that long for a new
drive or doing anything significant (like multiple shrinks!) to a partially
failed array sends shivers up my spine.
I may have to rethink my position on using raw drives. If I
partition the drives, I can make the partition a bit smaller than the whole
drive, allowing for the addition of a future drive whose size is a bit off.
I hate to waste space, but being stuck with an undersized or limping array
is worse.
Some manufacturers use the HPA (host protected area) to reduce the size
available to the user. You will see reference to this in the dmesg
output. There is a tool to let you see/set HPA, but I can't put my hand
on the info right now, the one I have is out of date, so I won't mention it.
--
Bill Davidsen <davidsen@xxxxxxx>
Obscure bug of 2004: BASH BUFFER OVERFLOW - if bash is being run by a
normal user and is setuid root, with the "vi" line edit mode selected,
and the character set is "big5," an off-by-one error occurs during
wildcard (glob) expansion.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
