[j-nsp] Problems with OSPF stub area (Juniper and Cisco)

Matti Saarinen mjs@cc.tut.fi
17 Nov 2002 12:09:18 +0200

Avram Dorfman <avram@juniper.net> writes:

> You're trying to get an ABR to "coalesce" (i.e. summarize, or
> aggregate, but coalesce is the OSPF RFC term) routes. In other
> words, you want to have multiple longer-mask routes in the range
>, in the stub area, but you only want to see the
> coalesced (or "aggregated") route outside the area.


     [ Sorry for the delay in replying ]

     Actually, I thought that wouldn't be the case but it is. There
     are multiple loger-mask routes in the area. The problem was that
     I saw _no_ routes in the area. Well, the /30 used on the p2p link
     did show up but not the others.

> So, I asked why you need such a route to be in OSPF in the first
> place. If you do, I suggested that you should somehow ensure that
> there really is a more specific route from the range in the area.
> Otherwise, you should be looking at either exporting a route into
> OSPF, or perhaps you are also running BGP, and you could deal with
> this route there.

     OSPF is the only way our router learns the routes in the area.
     Well, not the only way. There is always the possibility of using
     static routes but it's ugly.  

     At the moment, the routes do show up. The solution was to change
     the Cisco's configuration in a way that it sees the ATM
     connection as a point-to-point link, too. I do still wonder how
     on earth the neighbour adjacency became Full in the beginning
     when the routers saw different link types.

> Sorry if my use of the word "coalesce" is confusing. People often call
> it summarizing, but that's not accurate when discussing OSPF, b/c it
> refers to generating type 3 and type 4 LSAs for other areas, which
> does not imply aggregation.

     It's okay. I'm the one who should have understood. It's just been
     a long time since I've last checked the OSPF RFCs or the Moy's
     book. Now, I do remember the term.

     Thanks a lot again.

- Matti -