OSPF NSSA area

[Sam Preston]

Hi,

I have got an OSPF network (see attached diagram) in which I have a totally NSSA area.

Routers C and D (in the totally NSSA area) have got a connected route (the LAN to which Server S is connected) which they import to OSPF into type 7 LSAs. This is the normal behavior for totally NSSA areas.

Routers A and B are ABRs. They have interfaces in area 0 and in the totally NSSA area.

I configure this command on Router A and Router B :

summary address X.X.X.X X.X.X.X no-advertise

This command is used to prevent the regeneration of the type 7 LSAs coming from the NSSA area to type 5 LSAs in the backbone area.

My question is :

Suppose the link between Router A and Router C fails.

Suppose an IP packet comes from the backbone and wants to reach Server S. If the packet arrives to Router A (not routed via OSPF to Router A but via BGP, so the question is not how the packet reaches Router A), how will it reach Router B then Router D then the Server S ?

In short, the question is : if the link between Router A and Router C fails, how will Router A know the route towards Server S since the type 7 LSA for the connected route of the server is not regenerated into a type 5 LSA in area 0, and thus Router A does not learn that route from Router B...

Thanks a lot for your help !

Sam.

[Edison Ortiz]

In short, the question is : if the link between Router A and Router C fails, how will Router A know the route towards Server S since the type 7 LSA for the connected route of the server is not regenerated into a type 5 LSA in area 0, and thus Router A does not learn that route from Router B...

It won't since the summary address command will prevent that.

Router A won't be able to reach Server S via Router B under this scenario.

You need to remove the summary address command so the LSA could be seen as Type 5 on Router A.

[Sam Preston]

Thanks Edison for your answer.

So how should I proceed with such a design ? I think I can proceed like this :

- Make the NSSA area a normal area (let's say area 1)

- Setup a second back-to-back link between Router A and Router B and configure it into area 1 (keep the other link in area 0).

- Configure 2 different OSPF processes on both Router A and Router B so that the routes their learn from area 1 are not propagated into area 0.

- Configure a "default-information originate always" onto Router A and Router B to give a default route to area 1.

With such a design :

- If the link between Router A and Router C fails, Router A is still able to route towards Server S via Router B.

- The routes from area 1 are not exported to area 0

Do you think it is the good way to proceed please ?

Thanks.

Sam

[Edison Ortiz]

Port density an issue?

How about connecting C and D to A and B on a triangle design?

If C loses connection to A, A will receive the NSSA route from D.

[Sam Preston]

Yes, triangle design is not an option since I haven't got enough 10-Gbps ports for this. Anyway it would not be so different from adding a second back-to-back link between RouterA and RouterB in area 1. This is just about adding one more link in area 1 to secure the path because basically area 0 can't help for propagating the route. So let keep the links that I already have and just add a subinterface (VLAN) between RouterA and RouterB. What we just need is another IP link, not a physical link since we already have a fully redundant architecture at the physical level.

By the way I read in the documentation from another vendor that they have multi-area interfaces capability. In other words, an interface can belong to several areas at the same time. The goal is to avoid transiting through low-bandwidth links and to prefer high-bandwidth inter-area links rather than low-bandwidth intra-area links.

Do we have such configuration available here or not (in which case I will add a second back-to-back link) ?

If we don't have, maybe we could ask a dev for this, don't you think ?

Thanks.

Sam

[Edison Ortiz]

With the subinterface approach between Router A and B, you could make one subinterface be part of NSSA area and the other subinterface be part of Area 0.

Router A will receive the intra-area NSSA route during link failure from Router B.

What's your plan for reachability from the backbone area toward server S?