[Linux-cluster] Downing secondary interface if primary fails?

[Lon Hohberger]

On Wed, 2007-10-24 at 08:24 +0100, gordan at bobich.net wrote:
> On Tue, 23 Oct 2007, Lon Hohberger wrote:
> 
> >> Say a machine is a part of a cluster. It has eth0 for WAN accessm and eth1
> >> for SAN/cluster access. Say eth1 fails - is there a way to get eth0 to
> >> shut down as well, so the machine doesn't appear to be alive and thus get
> >> traffic routed to it?
> >
> > Use fencing to power the node off?
> 
> Can't power it off in software. :-(

No, it requires a remote power switch.  However, the remaining nodes in
the cluster will need to fence the dead node anyway (being cut off from
the cluster will make the node need to be fenced)...

Internally (i.e. local to the node that's been cut off), you could use:

http://mon.wiki.kernel.org/index.php/Main_Page

-- Lon