I am not talking about clustering here.
I have 6 hosts with internal IP addresses.
2 hosts ( group A ) have a single NAT'ed public IP address
2 hosts ( group B ) have a single NAT'ed public IP address
2 hosts ( group C ) have a single NAT'ed public IP address
All behind a firewall on the same VLAN.
What I intend to do is if one of the host dies, I add the IP address of
the host that died to the other host via a logical interface ... and the
services that were running on the host that died will be ran on the
remaining host.
It sounds simple:
When a host on group A dies, I can add the IP address of that dead host
on say a host on group B. All incoming TCP requests to the dead host
will be received by the host on group B, with the firewall not knowing
any better ( well, except for the MAC address ). If the incoming TCP
requests are load balanced between hosts on a single group, the firewall
would continue to load balance the incoming requests.
What is not so simple are outgoing connections:
When a host on group A dies, I can still add the IP address of that dead
host on say a host on group B. However, for specific TCP destinations,
some outgoing connections must originate with either the original IP on
the host on group B ... or the added IP address on the host on group B.
I could not figure out how to specify that specific outbound connections
must have a source address of x.x.x.x, while some with source address of
y.y.y.y
Any ideas ?
|
Similar Threads
Linear Mode
