|
Thanks, but I am only looking at failover.
I am trying to have a service continued should it fail,
i.e if the web server fails, then another machine takes over.
Matthew Marlowe wrote:
Well, the answer depends on what you are trying to do :)
The standard config we've used is to assign
one real IP for each box, and one virtual IP for
routing a virtual subnet of ips.
e.g., your upstream subnet is 10.1.1.0/29
10.1.1.1 upstream router vip
10.1.1.2 upstream router1
10.1.1.3 upstream router2
10.1.1.4 loadbalancer vip
10.1.1.5 loadbalancer1
10.1.1.6 loadbalancer2
Your actual websites, which would be on different
ip's due to different loadbalancing policies,
could be assigned 10.1.1.16/28
so you would have
route add 10.1.1.16/28 gw 10.1.1.4
statically configured on each loadbalancer, and
route 10.1.1.16 255.255.255.240 10.1.1.4
on your upstream router.
and, you would assign the ips however you wish,
10.1.1.17 first set of websites with same lb policy
10.1.1.18 second set of websites with same lb policy
.
.
10.1.1.30 fourteenth set of websites with same lb policy
|
