Claude Marinier writes:
> This is addressed to those of you who use NetBSD to do serious work. Here
> we use one as the principal DNS server. It will soon be a DHCP server. It
> may never be a mail server (the rest of DND is going to Exchange and we
> will probably have to follow suit).
> 
> When you upgrade, how do you proceed to minimize down time? Do you build a
> fresh system on a spare box and them swap? Do you use the upgrade option
> of sysinst? if so, how do you handle the modofied files in /etc?

I've upgraded various NetBSD boxen (including my nameserver/loghost/
ntp server box) from 1.3.x to 1.3.3 using something along the lines
of:
	* extract new kernel source
	* rebuild new kernel (i have a tweaked config)
	* mv /netbsd /netbsd.old
	* cp -p netbsd /
	* chown 0:0 /netbsd
	* cd /
	* get the 1.3.3 .tgz files into /var/tmp/133, except etc.tar.gz:
	* foreach i (/var/tmp/133/*.tgz)
	* gtar --unlink -zxpf $i
	* end
	* extract etc.tgz to a separate directory and compare (there
	  should be minimal (if any) changes to prior 1.3 releases)
	* reboot

Your milage may vary, but I did this and the machine had a minute or
so unavailibility during the reboot.