Two sites advertise same server address?

[j.vong]

Assume we have 3 sites network. Assume Site A (Client remote site) has a Frame Relay Port with two pvc going to Site B (Server hub site) and C (Backup server hub site). If Site B and C both have the same server address 10.1.1.1 since server in site C is a backup duplicate server of Server in Site B. What will happen?

Will Site C router sees both paths to 10.1.1.0 network and send traffic to both paths? Or will we simply crash the routers?

Thanks for all your help!

[Richard Burts]

What will happen depends on some things that you have not yet told us, but crashing the routers is hardly likely to be the result.

In part what will happen depends on what routing protocol you are running and how it advertises the server address. If the routing protocol advertises two equal routes then there will be two routes on the router at the client site. This means that traffic may go to one or the other or both depending on whether per destination or per packet load sharing is being done on the router.

Assuming that you want the traffic to go to the primary site if it is available and to the backup site if the primary site is not available, you would want to arrange the routing protocol so that the route advertised from the primary site was more attractive than the route fro the backup site. In a protocol like EIGRP it is fairly easy to arrange the metrics to favor the primary. In a protocol like OSPF it is more difficult but should be able to be done.

HTH

Rick

[j.vong]

Hello Rick,

Thanks very much for your help. Oh yes, I forgot to mention the routing protocol. It's EIGRP. And you are correct, I would like Site B be the Primary and Site C be the backup only when Site B is down. Can you share a few command on how Site B and Site C config should be in order to make Site B more attractive? (originally I thought I just set a delay metric higher in Site A on the subinterface pvc going from Site A to Site C). But then if there is a better way to set the preference on the Hub side (i.e. at Site B and C, instead on the Client side Site A), please let me know.

In that case, if we can set preference in the Hub sites rather than Client side, does that mean it doesn't matter whether the WAN network is layer 2 frame relay or layer 3 MPLS, and it will still function?

Thanks again!!!

[Richard Burts · Accepted solution]

The good news is that manipulating the metrics is fairly easy and straightforward in EIGRP where it not in OSPF.

You are correct that one approach that will work is to set the delay on the interface. But I prefer to use another mechanism. EIGRP supports the concept of offset list to manipulate metrics being advertised. I prefer this mechanism to changing delay because offset list gives more control and the ability to apply the change selectively. With offset list you can manipulate the metric of inbound updates (what you learn from a neighbor) or of outbound updates (what you advertise to a neighbor). And offset lists give you the ability to affect all routes being advertised or by using an access list within the offset list you can affect only certain routes. And with offset list you can affect any route regardless of which interface type is being used.

So you could define an offset list at site A which would make the routes advertised from site C less attractive. So the preferred route would be from site B and only if site B were not advertising would you use the route from site C.

Assuming that you want to affect all the routes advertised from site C the command would look something like this:

offset-list 0 in

where is the amount you want to add to the route from site C (you might need to experiment to find which value works best) and is the identifier of the interface (in your case the serial subinterface).

HTH

Rick