Routing

I'm very new to Junos and I'm a little confused about import and export policies. I just realized last night that they're from the perspective of the routing table and not the routing process, which certainly helped.  🙂

The OSPF default export policy is to reject everything, but since this is a link state protocol it still floods LSAs. If I understand this correctly, it has a couple of implications:

1. Single-area OSPF will work without configuring any export policy because all LSAs will be flooded.

2. Multi-area OSPF would require export policies in order for routes from one area to be advertised to other areas since this goes beyond simple per-area LSA flooding.

Is that correct? It's very confusing to read that the default export policy is to reject everything. In context of OSPF, I just can't figure out what it means.

Thanks,

John

hi Neiby

Before proceeding ahead, u need to understand concept of Backbone area first... This is the area to which all other areas are connected and is used to transport traffic/routing information in between multiple areas (if and when required)...

and then there are area area types you define while creating areas which define rules for routes to be passed in or out of that area , without requiring any additional export policy, definitly through LS Updates.... so here these  routes tell one area the routing information of other area.....(in an area u select DR which is responsible for intra area flooding of routes )..

Still u can use policy as and when u need to pass on some specific routes which are not being passed on otherwise(like some static routes etc)......

studying the above mentioned topics in detail will help u to understand better....

Hi John,

    Let me see if I can help you out here.

+---------+
| ROUTING |    <-------- OSPF/BGP/ISIS
|  TABLE  |    --------> OSPF/BGP/ISIS
|         |
+---------+

Anything the protocol pushes to the routing table would imply default-import.
Vice verse anything the protocol pulls from the routing table would imply default-export.

In sense, OSPF will never pull any routes from the routing table for advertising.
Basically the convergence of the Link State Protocol is based on the LSA/PDU.

In contrast, BGP will pull BGP routes from routing table for advertising based on the BGP algorithm.

However, when the user wants to export some specifics routes available in routing-table,
then export policy would be required for advertising these routes.

Hope this helps.

Regards
Surya

That explains my confusion. Even though I know the OSPF database and the routing table are quite different, I was subtly mixing the two concepts when trying to understand OSPF and import/export policies. When you said that OSPF doesn't need to get routes from the routing table to export because it is "exporting" from its LSA database, I saw the source of my confusion and it all makes sense now.

Thanks!

   Greeaaaattt.

   You will have ospf routes in your forwarding table, only if this have better preference than other routes There is no way, or you should not filter routes in a ospf domain. This wil create several topology or ospf database, and will lead to inconsistent routing or path decisions.... so routing problems.

  Filtering usually only works for external routes, and or between ospf areas, never inside an area, or works from other routing protocols inside ospf.

  Only keep in mind, every ospf Router takes its own routing decisions, based on its ospf database. There is no route flooding like in other protocols.

   Br

   Alex.

  Hi Neiby.

  Export and imports could be quite confusing. First, think that a router should propagate only accurate routing information. This is the main point to flood information from the forwarding table instead of information from the routing protocols. A router should inform other routers only about active routing information, this avoids loops and wrong path selection. And only advertise in each protocol the active routing information for that protocol.

  But OSPF is quite different to other routing protocols.

  Under OSPF, you dont have to  configure any export policy, because all neighbors should have the same OSPF Link State Database, i dont said route database.

  OSPF is a link state protocol that uses the Djistra algorithm. This use a topological database for path or route selection. In order to avoid loops and wrong route selection, all the ospf routers under the same domain (not area) should have the same database, this is because you can't filter ospf internal routes, so the default import or export policy is a "flood all". All routers should have the same network view, diagram, topology.

  This ospf database is not a route database, is a topology or "network diagram". If there is a link failure, all routers have the topology knowledge to select another path across this topology database, so dont have to wait for a route update like other routing protocols.

  ospf use a information unit called lsa for information flooding, not route flooding.

  Under a link failure, there is only a lsa update about the failing link, so every router know how to react to this failure and select other path if available for every destination network across that link.

  This way the network dont have to flood new routing information for ¨1000¨ routes, only have to flood the updated lsa and every router compute the changes.

  The others protocols, RIP, EIGRP, BGP, support route filtering, because they dont need a complete network view, and this view is not a network topology. So you could manipulate its routing information as needed, this is why these use export and import policies.

   ISIS is near like OSPF, so it also uses a default flood all.

   RIP export policy is reject everything, have to configure a export policy to flood routing information to other neighbors.

   BGP default is flood all.

   The default import for all is import everything. After import the routing information, the router have to go through a route selection process, selects from all the routing protocols the best routes, using best preferences values, longest matches, etc, and build the forwarding table.