Adding (removing) disks to a RAIDZ means redistributing all the data in the pool
across the new set of drives (adding a drive, or freeing up a drive if enough 
space
is available).

That would be difficult to do correctly, especially since it would be slow 
enough
that most folks would want to be able to continue using the pool while that was 
done,
and without too much performance impact.  So it probably has to treat itself as 
if
it's in at least two parts (old layout and new layout) plus whatever it's 
working on at
the moment.

I think for example Veritas Volume Manager can do something a little like that,
re-arranging storage in place.  I haven't looked recently, but it wouldn't 
surprise me
if even that has limitations as to the sorts of in-place rearrangements it can 
do.  And
in my experience, while VxVM can be very good when everything works, it's rather
complicated to use and an absolute nightmare when things don't work; and it can
sometimes be less robust in the face of hardware failures than one might hope or
reasonably expect.  And of course VxVM isn't free (well, there _have_ been 
scale limited
free promos from time to time).

Nor, beyond a certain point, would it be a good idea to add drives to an 
existing
RAIDZ even if it were possible.

AFAIK, you can:

* replace all the drives, one at a time, with larger drives
* add an entire 2nd RAIDZ (i.e. a set of three or more drives) to a pool

Given a little foresight (and unless one is trying to do stuff on the extreme 
cheap),
how many situations are there really that _need_ the additional option?  Even if
there were a "bounty" or some market mechanism like it, would there be enough
demand to pay for not only the initial development, but also the _maintenance_
of such a capability?  "bounties" may be a useful concept (although more 
difficult
in a regulated corporate-sponsored environment than in a 100.00% 
community-driven
environment, which is why I wonder if they will ever happen except in an 
isolated adjunct
process), but I don't know that this would be a good test case for them.
-- 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to