>>  Maybe it's a silly question, but is it possible to add one or more
>>  disks to a ccd partition without damaging its existing files?
>
>Yes it is.

So, how would one do it?  Perhaps by...

- creating an additional ccd device from the new partition(s)
- newfs'ing it
- unconfiguring the new and original ccd devices
- adding the new partitions to the original device's line in
   /etc/ccd.conf, and finally
- configuring the new, expanded ccd device?

But that doesn't sound right to me.

>>  I guess, if it's possible at all, it would only be so for partitions
>>  which have been concatenated with no striping?
>
>No; it's just _easy_ only for ccds which are concatenated rather than
>striped.
>
>For a striped ccd, you'd need to work out what blocks need to go where
>to preserve the virtual disk's contents, and move them.  I don't know
>of any extant tool which does that, but you said "possible", and it's
>certainly possible to build one.

OK, I'll forget that for the time being, although I'd like to know 
how to go about writing such a program.

>>  By the way, I seem to remember Jason mentioning somewhere in the mail
>>  archives that there are no benifits to be gained by striping IDE
>>  disks, yes?

I managed to find the message I was thinking of:

   http://mail-index.netbsd.org/netbsd-help/1996/11/24/0001.html

in which Jason said

   "Unfortunately, with IDE, you don't get _any_ concurrency.  Quite honestly,
    with IDE, you'd be better off with pure concatenation (interleave == 0)."

Is that the case even if only one disk is attached to each IDE 
controller, or only for a ccd made from two disks, both on the same 
controller?

I'm thinking of making a ccd from seven or eight IDE disks so I'll 
need some PCI IDE controllers, any suggestions for ones supported by 
NetBSD?  I wonder if noticeably faster disk I/O speeds would be 
obtained by using PCI cards which have only IDE controller each 
instead of two, since two controllers on a card might create a bottle 
neck?


Ray