If the boot-blocks are already booting properly, then they should not need =
to be changed. You can manually install a new root partition in raid0h (or =
other free space).

Then put the new kernel in raid0a, eg. nbsd16.new (same location as 1.6K)

At the boot console in interactive mode (man boot)
 boot nbsd16.new -a
The new kernel should load and will ask for location of root/swap: =
raid0h/raid0b.

This is a manual process to boot the new version, but you could build a =
new kernel with raid0h hard-coded and replace the original 1.6K, now =
called netbsd.old.
Then to boot to the old version, simply choose netbsd.old.

>>> Jeff Rizzo <riz@tastylime.net> 04/12/05 13:15 PM >>>
I think I'm having a problem of imagination here, but I'm trying to
figure out a way to boot two different versions of NetBSD from a single
logical disk which is a raid1 mirror (using raidframe).

What I have:

a PC-type system with serial console (and no remote BIOS access) in a
colo 50 miles from where I live
two hard drives in a raid-1 configuration (and no way to easily add =
drives)

This system is currently running NetBSD 1.6K, and I'd like to upgrade
the system to something newer (ipf 3.4.29 is giving me trouble) , but
provide a quick and relatively painless (as in, does NOT involve a drive
to the colo) way to revert back to the previous system should there be
unexpected problems with the upgrade.  (After some period of time, it
could also provide me a way to upgrade to something newer still, should
it prove necessary)

The stumbling block I'm running into is how to _boot_ the new system.  I
attempted to add a second root partition at raid0h, duplicating the
size/offset info (plus RAID header) in wd0h, and specifyin that to the
bootblocks, but it doesn't seem to work.  I'm open to other suggestions
(including creating two RAIDs on the disks, should that prove
necessary... that's kind of a pain, though, since i'm already
partitioned and installed on the single-raid setup.

Thoughts and suggestions gratefully accepted.

Thanks,
+j