Frame-Relay Hub and Spoke

abbas.ali · ‎05-31-2003

I have a r1 (hub) and r2 and r3 (spokes) are connected as network type non-broadcast which is default on physical serial links.

I am running ospf backbone 0 and advertising a network on r2 such as 100.0.0.7. r3 sees this network in it's routing table reachable via next-hop of r2 instead of nex-hop to r1. Is it normal.?

This is hub and spoke and there is no pvc between r2 and r3. r3 is learning this route ofcourse via r1 (hub) but the next-hop is not changing.

I always thought that this kind of behavior only applies to bgp internal peers.

How do I fix it if it is not normal?

Thanks,

dnagarajachary · ‎05-31-2003

Can you reach r2 from r3? that is can you ping from r2 to r3.

What does traceroute from r2 to r3 give?

could you give the configurations details?

-Deepu

ruwhite · ‎06-01-2003

The second hub should see the route with a next hop of R1, the hub, rather than R2, the other remote site. Could you post the configuration for the three routers (we don't need all of the configs, just the serial interfaces, and ospf configurations), a show ip route from the three routers for one route that's attached to the second remote site, and a show ip ospf data for the same route?

Russ

mark-obrien · ‎06-01-2003

In the default condition, with the serial interfaces set as non-broadcast, the OSPF process will assume that each router can reach all the others directly on the non-broadcast medium. This is why R1 is not showing up as the next hop between R2 and R3. In your case, the routers can not reach each other over the non-broadcast network. For OSPF to work in the non-broadcast mode, you either need a full mesh on your frame relay, or manually map neighbors to DLCIs. This is why the preferred methods for running OSPF on a non-broadcast multiple access (NBMA) network are to either configure the interfaces as point-to-multipoint or to set up point-to-point sub-interfaces.

HTH

Mark

rjackson · ‎06-02-2003

In ospf routers forward each others advertisements and each router builds a map of the network and chooses the best path to use. In other words r1 is not advertising r2's networks to r3 and vice versa, it is forwarding their advertisements between them. r2 and r3 are making their own decisions about the next hop to reach a destination independent of what r1 is going to do. You MUST be using multiple map statements in r2 and r3 or this wouldn't be working in the first place. So r2 and r3 each have map statements that tell them they can reach each other directly so they build their routing table accordingly. R1 is providing a transparent link between them.

dnagarajachary · ‎06-02-2003

OSPF handles Multipoint frame-relay interfaces as Non_broadcast by default. When it considers the interfaces as non broadcast, then the requirement is that, the neighbors should be explicitly configured.

If you are having the interfaces as multipoint, then you need to explicitly define

the link as either point to point or broadcast

if you have point to point link,

then under each multipoint interface, give the command

ip ospf network point-to-point.

then the routers would see the proper routes being exchanged.

-Deepu

mark-obrien · ‎06-02-2003

There is another possibility. The NBMA network does have a DR. Either by luck or design R1 became the DR and passed R2's and R3's LSAs on. The link state database shows the NBMA as a transit network on the path between R2 and R3. In constructing the routing table, the routers showed the frame relay interface address of R2 and R3 as the next hop towards the networks on those routers, as they would on any transit network. Whether R2 and R3 can reach each other depends on whether there are multiple map statements, as RJackson describes. Output from the "debug ip packet" command will show encapsulation failures if these map statements are not present.

Mark