>>>   disklabel: Invalid signature in mbr record 0
>>>   write: Read-only file system

> If I understand your suggestion correctly, you recommend that I move
> raid0e forward a bit to clear the raid0 disklabel so that the cgd on
> raid0e won't stomp on it?

I did that, and it worked!  Thank you very much, Lubomir.

For anyone reading the list archives and hoping for more detail, 
where I had this on the underlying device (raid0 in this case):

   c: 480203953        63     unused 
   d: 480204016         0     unused 
   e:  10485760        63        ccd 
   f: 469718193  10485823     4.2BSD 

I changed it to this:

   a:        16        63     unused
   c: 480203953        63     unused
   d: 480204016         0     unused
   e:  10485744        79        ccd
   f: 469718193  10485823     4.2BSD

In the above case, dummy partition "a" reserves space for the
disklabel, so that it will not interefere withe the disklabel of the
cgd (on partition e), which is used for cgd verification.

Thanks also to der Mouse, who suggested in private communication that
16 sectors (8K) would be an appropriate size to skip, since it
corresponds to the amount skipped by FFS to leave room for boot blocks
and disk label, so this should be enough.

Now on to my next cgd issue, about which I'll post under a new thread
if I cannot find an answer via Google...

Anne.
-- 
Ms. Anne Bennett, Senior Sysadmin, ENCS, Concordia University, Montreal H3G 1M8
anne@encs.concordia.ca                                    +1 514 848-2424 x2285