Good day folks,
Here's a tough one, I think at least, can anyone give me a hint here?
At my client's site, hosts on a single-subnet LAN on network 10.x are set up with a PIX 501 firewall as default gateway. This device doesn't support multiple default routes, so to introduce redundancy in connecting to the Internet, I had to introduce a W2KP PC as a router (with IP Routing enabled in the registry) and set this PC up with two default gateways. Gateway 1 has metric 1 whereas the Gateway 2 has metric 2.
The W2KP PC correctly selects Gateway 1; when I unplug the UTP cable to Gateway 2 it neatly fails over quickly onto Gateway 2. When Gateway 1 is re-connected, it quicklyre-selects Gateway 1 as default gateway (checking default gateway using route print).
My problem is this: I approached the design problem as a dead-gateway-detect case wherein the "host" at one end of the TCP connection selects the next gateway in its list after waiting in vain for acknowledgements to data segments for half the number of times in the TcpMaxDataRetransmissions registry setting. On second look, it seems to be the wrong approach because my W2KP device is not at the ends of a connection but simply examining destination IP addresses in its router role.
So is the elegant fail-over working simply because W2KP is detecting that the UTP cable to the respective gateways is getting detached?