[f-nsp] Getting started with ServerIron 4G

[Chris Price]

Oliver Adam wrote:
> The virtual server failover is bound to the switch failover in a 
> hot-standby setup. There is no need for the real servers to fail over. 
> They do see the same MAC before and after the failover and I do not 
> know what you mean with a real server failover.
>
     I am obviously missing something easy here. I'll try and illustrate 
my simple setup that I am using for learning purposes with the SI's;

    I have two SI's: lb1 and lb2 configured in a hot-standby ha pair. I 
defined a 'server source-ip-standby <ip> <mask> <gateway>' on lb1 and 
lb2 to provide a failover gateway ip for systems being load balanced. I 
have tested the failover gateway ip functionality and it works as expected.

   Lets assume that lb1 is my current active node and lb2 is the standby 
node in my hot-standby ha pair.

   I defined two real servers (realA and realB) and a virtual server 
(virtA) on lb1. I bound realA and realB http port to virtA. Http 
requests to virtA's ip sees traffic being distributed to both realA and 
realB in a round-robin fashion. This is all expected for my basic 
configuration.

THE PROBLEM:

When I force a 'reload' on lb1, and the hot-standby HA configuration 
fails over to lb2 (ie: lb2 becomes the active node), my real servers and 
virtual server config are not 'failed' over.

My experience with Cisco Arrowpoint and F5 is that there is some sort of 
sync to peer config that must be run to sync config changes on the 
active node to the backup node. For the life of me I cannot find any 
information in the Foundry docs (slb guide, hardware install guide, etc) 
on how to propagate changes from the active node to the backup node.

Does that make sense?