On Wed, Apr 17, 2002 at 08:14:55PM +0100, tld wrote:
> (should this go to netbsd-help?)
> 
> Hello.
> I am in need of developing a High Availability solution to my innermost 
> problems, and I'd like to ask a few questions.
> 
> The goal may sound as follows:
>    With the least cost (recycled hardware :) ) produce an HA server 
> possibly secure and with load-balancing. The so-called server cluster 
> can be connected to the rest of the net via a single network cable and 
> has a single network address. IPv6 support is a welcome addition :)

You might want to take a look at the freevrrpd(*) project.  It will allow
you to fail a vip between machines.  It still sounds like you'll have
plent of SPOF issues, so that is perhaps not worth the effort.
If you get a couple of switches that support STP and cross link them, plus
have multiple uplinks you can have redundant network connectivity, then
the freevrrpd solution would be useful.  You still need to fix the backend
SPOFs, so again unless you're going to build a completely redundant system
you may not want to put in the additional work of porting freevrrpd.  You
might just some homegrown scripts to poll the upstream router and take some
action if you lose connectivity.

So, freevrrpd may solve your frontend network failover, but that is a
small piece of the HA issue. 

-Andrew

(*) It is currently FreeBSD only.  See http://www.bsdshell.net/hut_fvrrpd.html