>Another pmax sysinst lossage?  Could you show us disklabel output of
>rz5 and/or rz6?  Because Jonathan spent much efforts to port sysinst
>for NetBSD/pmax one year ago (I was surprised how much NetBSD/pmax
>kernel was changed to achieve that), he might know better.

I think its probably that the 1.3.2 MAKEDEV script was not creating
device nodes for rz5 or rz6. This is fixed in 1.3.3.

 If you can get the root filesystem mounted readwrite at all -- via,
e.g., mouting kernfs on some handy directory, like /kern (or /mnt, if
that's not there) then all is not lost:

       mount -t kernfs kernfs /kern
       mount /kernfs/bootdev /

(or mount -t kernfs kernfs /tmp; mount /tmp/bootdev /)

then create the device entries for rz5 and rz6:
     cd /dev  
     /bin/sh MAKEDEV rz5 rz6

At that point it might be simplest to reboot and start again; i have
no idea what mounting the root via kernfs/bootdev will do, tho' i hear
its supposed to work.  Or you could just do the MAKEDEV on the NFS server.

Just to be sure: you are installing 1.3.2 or a recent -current, right?
the changes to support scsi disks at units 5 and 6 weren't in 1.3.0 or
1.3.1. 

Can you send the dmesg output?