Subject: Re: wedges vs. not-quite-wedges, was > 1T filesystems, disklabels,
To: Bill Studenmund <wrstuden@netbsd.org>
From: Dr R.S. Brooks <R.S.Brooks@liverpool.ac.uk>
List: tech-kern
Date: 12/20/2002 13:12:31
On Thu, 19 Dec 2002, Bill Studenmund wrote:

>
>I mean, come on. What RAID system wants to expose its individual
>components to userland? The completed RAID sets, yes, but not the
>components.

What about when a compnent has failed, you've fixed (or replaced) it,
and you want to tell the RAID system to regenerate the data on the fixed
component.

And as an aside this gets even messier with fibre channel disks where
every disk has a unique WWN.  In this case you have to be able to say
"replace the disk with WWN foo by the disk with WWN bar (and if it matters,
you will NEVER see foo again, because it is dead and there will never be
another disk with the same WWN)".

I'm not sure I've quite "got" the concept of wedges, but one thing bothers me.
If the "partition" information is stored somewhere else (other than on each
disk), how easy will it be to handle the situation where you want to
rearrange a large disk farm and preserve all the data?  Say you add more
controllers and want to spread the existing disks over more SCSI busses
without having to dump all the data to tape and restore it again.


Roger