[Rack] Monkeybrains link seems to be dead.

[Dr. Jesus]

On Tue, Jun 22, 2010 at 5:39 PM, Matt Peterson <matt at peterson.org> wrote:
> Still happy to upgrade the Soekris image.  If someone has cycles to
> reconfigure the AP's & switches to be trunks, we can with the original
> plan of beaconing out 'noisebridge-sonic', 'noisebridge-monkeybrains',
> etc; let the user decide - not us; thoughts?

That's one way to let the users get a failover option.  Here's another
with different advantages/disadvantages: carp + pfsync + two NAT
pools.  The two soekrises would share a carpX interface on
172.30.0.0/22, and carpY and carpZ interfaces on 192.168.1.0/24 and
192.168.188/24.  Hosts would randomly be assigned to a NAT pool
address on one of the two subnets using the usual CARP load balancing.
 That's what I meant by an aggregation network in my earlier email
today.

Failover behavior:

 - When ifstated detects that one of the WAN links dies, we can
configure it to take the appropriate action wrt the default route on
both gateways at the same time.

 - When one of the soekris devices dies, the remaining partner will
continue to process traffic and hash flows to one of the two NAT
pools.

It would be super nice if we could also load balance our transmitted
traffic on both upstream links simultaneously, but I have no idea off
the top of my head how to implement that in OpenBSD with OER-like link
characteristic compensation.  I might check tonight for ingress
filtering on MB and sonic's gateways, though, just to see if that's
even possible.