I have several FreeBSD 6.0-RELEASE servers connected to a local area
network in Atlanta.  The network design is setup so that there are two
pairs of routers, one at each end of a T1.  Each end of the network uses
GLBP (Cisco's Gateway Load Balancing Protocol) to distribute load over
the two routers, and provide redundancy in the event of a failure.  This
design works great when there are lots of hosts connecting to the glbp
interface, because GLBP provides a virtual address that the two routers
share, and then alternate arp responses between the two routers.

The problem I am having is that each server cache's the mac address of
router A or router B for at least 5 minutes.  This means that for 5
minutes it sends it traffic to router A, then for 5 minutes to router B,
and so on.  I need a way to tell each host to re-arp a lot more often.
Or, maybe I am thinking about this the wrong way.  

The end result, I need a way for the server to distribute the traffic
over the two gateways.  I tried using IPFW's to write a rule that I
found online to use probability (50%) to fwd the outbound packets to
another IP address, but that didn't seem to work.  If this was possible,
I could easily setup two GLBP interfaces on the lan and we could have
the FreeBSD server alternate outbound packets between them.

Any comments or suggestions would be greatly appreciated.

Thank you,

freebsd-questions@freebsd.org mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to