[Linux-HA] advice on configuration

[Marshal Newrock]

On Sun, 14 Jan 2007 11:33:54 -0700
Alan Robertson <alanr at unix.sh> wrote:

> Marshal Newrock wrote:
> > Since I'm not sure what I'm planning will work as I intend, I'd like
> > some advice on how to set up my cluster correctly.
> > 
> > This is a two-node active-active cluster.  If services on one node
> > fail, or if it loses its internet connection, the services should
> > all migrate to the other node.  If the heartbeat communication
> > between the two nodes fails, then it should ignore all other
> > failures until there's a good link again.
> > 
> > What is the correct way to do this?  Links to any documentation are
> > welcome.
> 
> This won't do what you want it to.
> 
> If the other node fails, it is indistinguishable from failure of the
> other node.  This means that if you have a node fail, that it will do
> nothing.  If that node was providing all the services, then that would
> mean that no services would be provided.
> 
> Not probably what you had in mind.

What about the case where one node loses its network connectivity on
the public interface, or has a hard drive failure, or the like?  Is node
failure still indistinguishable from node failure?

-- 
Marshal Newrock
Ideal Solution, LLC - http://www.idealso.com