Routing

Hello everyone,

i have folowing setup:

MX80-----SRX1400-Cisco

I am going to introduce OSPF with 2 areas.

I have read i need rib groups or export/import policies  to do that, but what about two ospf areas on SRX1400 ??

I will have area 0.0.0.0 towards Cisco and another area 100 towards MX80.

Obviously both devices MX80 and Cisco will participate in ospf`s processes.

My aim is that SRX1400 will update Cisco and MX80 about routes in both mentioned areas and additionaly if possible route summarization will take place .

Do i need some policies or anything else to make this working???

I am asking becuase i don`t know what i can obtain having only configured ospf and two or more areas in other words when reas are enough but when i need additionally rib-.. or routing instances...

By default, OSPF propagates OSPF learned routes.  Areas do not filter LSA advertisements by default.  To introduce routes that are not native OSPF routes, like statics, BGP or RIP, you will need to configure an export policy that matches the routes you want to "redistribute". 

You cannot configure an import policy on OSPF (a more involved topic for a later time)

You do not need to configure RIB groups to "redistribute" routes.

I'm attaching a redistribution primer I did for my co-workers (Juniper SEs), and there is some poking fun at Cisco (tasteful).  There are configs included in the presentation.

Attachment(s)

route redistribution.pdf   708 KB 1 version

Thank you very much -probably you have hit the home.

I have gone through the pdf and i have few questions.

1) OSPF/ISIS -Reject everything. (The
protocol uses flooding to
announce local routes
and any learned routes.)

Should i understand by "everything" everything except OSPF has learnt by itself ??? This is not so clear for me

2) In routing policies i see protocols direct and static - if i want to export static routes whiach one should i use ??

What is the difference ??

1 - Yes - OSPF will propagate any routes based on the I/F's in the OSPF area and it will propogate any routes learned through OSPF. As the example shows you can create a policy to capture a route and the use the export command to push that route into OSPF.

2- If you want static routes specify static - direct refers to directly connected routes. If you look in your routing table you will two entries for each route that is reachable via the I/F IP. Direct and Local. Local being the I/F IP address itself and Direct being the associated directly connected route. Static being routes you define under your routing-options section.

And you can of course also create a policy for direct. However by default any I/F in an OSPF area will have the direct route included in OSPF (unless you set the I/F to passive)

Really good explanation ! -Thank you.

Small example:

On the router i have 3 interfaces(networks) : ge-0/0/1, ge-0/0/2,ge-0/03

Ospf enabled is only on the ge-0/0/1

Now, by default my OSPF`s neighbor (via ge-0/0/1) will get all routes learned by OSPF and route directly connected to ge-0/0/1

If i would like to share with OSPF`s neighbor information about networks connected to ge-0/0/2 and ge-0/0/3 i should

use polic and export direct routes to OSPF, then it will send info about all directly connected routes, am i right ???

You could also include ge-0/0/2 & 3 in your OSPF definition and set them to passive and they would be injected into OSPF but would not have any adjencies formed on their interfaces.

IE -

user@host# set protocols ospf area 0 interface ge-0/0/2.0 passive

There are two ways to do what you want in your example.  The way you described it works, but you introduce the other interfaces (ge-0/0/2, ge-0/0/3) as external OSPF routes.  In Junos, external OSPF routes have a higher preference (150) than internal OSPF routes (10).  OSPF also prefers internal routes versus external routes.

My suggestion is to add ge-0/0/2 & ge-0/0/3 in OSPF as passive interfaces.

[edit protocols]
jparks@j5# show 
OSPF {
    area 0.0.0.0 {
        interface ge-0/0/1.0;
        interface ge-0/0/2.0 {
            passive;
        } 
        interface ge-0/0/3.0 {
            passive;
        }
    }
}

This keeps them as internal OSPF routes, yet prevents hellos and adjacency formation out of those interfaces.

Ok, but here is written like that:

"And you can of course also create a policy for direct. However by default any I/F in an OSPF area will have the direct route included in OSPF (unless you set the I/F to passive)"

Based on what is written above for me obvious was that passive interface is not included in OSPF LSDB and thus in updates sent to OSPF`s neighbor.

Have is missunderstood something ???

Remember that OSPF uses a LSA Database (LSDB) to flood information between OSPF neighbors.  This is different than a route-table.  When you specify an export policy, you are exporting routes from the route-table to the LDSB - and by default no routes are sent from the route-table to the LSDB (hence the reject-all).

When you add interfaces under the "protocols | ospf " stanza, these interfaces get added to the LSDB as Type 1 and NBMA networks get added as Type 2 LSA.  This is how OSPF initially populates the LSDB.  These interfaces become"native" to OSPF.

Understanding Junos policy is about centering your mind on the route-table (RIB) and think about importing into the RIB or exporting from the RIB.  The RIB is the reference point for everything.  Next you have to take into account how the protocols work.  With BGP, the protocol simply advertises the prefixes to the neighbors (connected via TCP).  But with OSPF/ISIS, the LSDB is what contains the prefix information, so you must think about the relationship between the routing table and the LSDB.  If you put routing information into the LSDB then that information will be flooded throughout the OSPF topology.