LVM + RAID follow up

Rich Braun richb at pioneer.ci.net
Tue Nov 7 19:44:14 EST 2006


"Derek Atkins" <warlord at MIT.EDU> wrote:
> I can't be the first person to think about this problem but
> not want an infinitely increasing number of MDn RAID arrays.

I think you're expressing concern about something that's really not as much of
an issue as you think.  If you did an average of one hardware upgrade each
year, you'd have /dev/md0 through /dev/md9 at the end of a decade.  By then
the whole approach to the problem will almost certainly have evolved, and the
tools you seek may very well have developed.  With today's tools, yes, in
order to shrink down the number of md[0-9] devices back toward 1 or 2 you'd
have to make a big fs on a new volume and rsync it, might be a few hours of
shuffling files.  But that's only after quite a few years of continuous
runtime.

dsr wrote:
>> Each partition for RAID must be the same size.
>
> Odd..  According to a modern mdadm man page (and a great LVM-RAID
> resource [0]) each partition does NOT have to be the same size

You're both right but the end result's the same:  the software will work best
if you make your RAID partition elements for any given /dev/md[X] device all
the same size.  Not so long ago, you had to use identical-sized drives
operating at identical speed:  things are much more flexible today.  I can
only assume system tools will eventually get better.  My point in bringing up
this whole topic is that today's tools are really quite nice but very few--not
enough--people use them under Linux.  And they aren't available at all on
Windows XP Home Edition, which is one reason I'm eager for further evolution
of the Linux desktop.

I hope that more of y'all decide to take advantage of RAID+LVM during your
next system installation or upgrade.

-rich


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.




More information about the Discuss mailing list