[Condor-users] Complex networking configuration

["Doherty, Peter Charles" <Peter_Doherty@xxxxxxxxxxxxxxx>]

I'm trying to set up a new condor instance in a somewhat tricky network
configuration.
This is actually for use as a glideinWMS frontend.  I'm hitting some
snags, so I thought I'd float it on the list, and see if anyone can make
sense of it.

I want to run a Schedd, Negotiator, Collector all on one machine, and use
a single configuration file, so they are all children of the same master.

The public IP/hostname routes to a load balancer, which forwards traffic
for the Collector ports (I'm running 5 collectors) to an internal
interface that is bound to the physical machine where the Central Manager
runs.
The CM has a public interface, that's firewalled, a private interface
that's nat'ed, and a loopback alias that's configured with the same public
IP that's on the load balancer.

Writing all this out, makes me realize that no one is going to be able to
help here.  What's the most obtuse network configuration that someone is
currently using out there, and how did you configure your CONDOR_HOST,
NETWORK_INTERFACE, BIND_ALL_INTERFACES, TCP_FORWARDING_HOST,
PRIVATE_NETWORK, etc...to make it work?
I swear I've tried every combination, and we already changed the network
topography once, but I can't make it all work.
Either the collector works, or the schedd works, or the collector is
accessible from offsite, but not the lan, or the schedd listens locally,
but not on the lan, or some other combination like that.

--Peter