How can I make a specific OSPF ABR as the default gateway for the totally stubby area?

[robert.huang]

Hello,

Please see the attach screenshot for the simple diagram.

Area1 is configured as the OSPF totally stubby area. R1 and R2 are both the ABR's. So both of them will send a default route to R3. The design is not good but I can't change it at the moment. I've configured "auto-cost reference-bandwidth 1000" on all 3 routers. Now I'd like R3 to always take the default route from R1, not from R2. How can I do that? I've tried to set "ip ospf cost 1" for Fa0/1 on R1 but it didn't help.

Please help.

Thank you very much.

[alexchev · Accepted solution]

Hi Robert,

If I understand your correctly, you just want to receive the default route into totally stub area from R1, but not from R2. I think, you can easily do it using LSA filtering on R2. I don't know what is your certain goal and platform you are trying to do it, but look, please at this:

http://www.cisco.com/c/en/us/td/docs/ios/12_0s/feature/guide/fiarospf.html

As soon as ABR should advertise only default route into totally stub area in LSA-3 updates, you can filter (deny) everything on R2, thus you will see all OSPF updates being originated from R1.

Tell me, if I'm wrong with the initial understanding of your question.

Thank you.

[Rolf Fischer · Accepted solution]

Robert,

not sure if this meets your requirements but you can specify the cost of the default (summary) route an ABR sends into a stub area with the area default-cost command:

http://www.cisco.com/c/en/us/td/docs/ios-xml/ios/iproute_ospf/command/iro-cr-book/ospf-a1.html#wp1610258897

HTH

Rolf

[Jon Marshall]

Hi Rolf

Haven't spoken to you in a while, hope things are good.

I'm still reading all your OSPF posts :-)

Just out of interest I saw this thread and instead of using the simpler solution you posted I tried out another solution of yours (for a slightly different issue) and used point to multipoint connections and then used a neighbor cost to make R2's default route a higher metric.

I also modified the hello timers because of the defaults.

It worked as expected but what are the implications from an OSPF point of view of doing this ?

In addition I am assuming it only works if there are no other devices on the common IP subnet other than the routers themselves ?

Jon

[Rolf Fischer]

Hi Jon,

always good to hear from you, I hope you're doing well, too!

As you know, the OSPF network types differ in the way they perform neighbor discovery/maintenance, LSDB synchronization, representation in the LSDB and, of course, the required characteristics of the underlying link technologies. The developers perhaps overshot their mark with all those options they gave us ;-)

In short, the Point-to-Multipoint is the better choice when the connection type is somewhat unreliable or direct communication among all host on the subnet cannot be guaranteed (or is even partially impossible).
On an ethernet segment, if host A can talk to host B and host B can talk to host C, we can be certain that host A can also talk directly to host C. This kind of connectivity characteristic is absolutely mandatory for Broadcast and NBMA to work properly, since every router must be able to communicate to the DR and BDR. The advantage of this network types is that they can handle a large number of routers on a common subnet efficiently in terms of database- and network consumption.
In a point-to-multipoint network, there is no DR/BDR. In the LSDB it is represented like numbered point-to-point links (*), with only one difference: Instead of the network ID and its mask, the interface ip address with a /32 mask is advertised in the second (stub) LSA. The result is that routers install host-routes to all its neighbors in order to overcome reachability problems on non-transitive links.

Traditionally, the decision between Broadcast/NBMA and Point-to-Multipoint (in other words DR vs. Non-DR types) was rather associated with technologies like Frame-Relay; however, nowadays it might be something to consider as well when using carrier ethernet technologies.

As you've  indicated, the IOS option of assinging costs on a per-neighbor basis on a common subnet with point-to-multipoint interfaces is sometimes interesting from a design perspective.

I may overlook something but I don't think that different network types affect the presence of other devices on the segment somehow.

I'm afraid I didn't tell you anything new, we can try to go into some of the details if you like :-)

Rolf

(*): Thats why both network types are compatible when we adjust the timers.

[Jon Marshall]

The developers perhaps overshot their mark with all those options they gave us ;-)

It's a constant source of confusion for me :-)

Please ignore the question about other devices on the subnet, particularly stupid of me.

I am revisiting some key protocols and at the moment am looking into multicast in a bit more detail so I am not thinking straight !

I just really wanted to know if there could be any adverse effects by using that solution and I guess not although your solution is a lot simpler.

Paul has added another one of which I wasn't aware, so perhaps a revisit of OSPF is on the cards as well if my brain can cope with it :-)  

Jon

[paul driver]

Hello Jon

I find it a constant battle is remember stuff -( a nice one) but  its groundhog day for me every couple of weeks!

res

Paul

 

[Jon Marshall]

Paul

I know exactly what you mean.

The worst one for me is when I am posting an answer to a question and do a quick search to verify it, find an answer on CSC and then discover it was my own answer !

This has happened to me more times than I care to admit :-)

Jon

[paul driver]

Hello

I managed to do this via adjusting the metric of the summary LSA on the least preffered ABR
 

router ospf 1
max-metric router-lsa summary-lsa 200


 

sh ip ospf database summary

            OSPF Router with ID (10.10.10.3) (Process ID 1)

                Summary Net Link States (Area 1)

  LS age: 421
  Options: (No TOS-capability, DC, Upward)
  LS Type: Summary Links(Network)
  Link State ID: 0.0.0.0 (summary Network Number)
  Advertising Router: 12.12.12.1
  LS Seq Number: 80000002
  Checksum: 0x56F9
  Length: 28
  Network Mask: /0
        TOS: 0  Metric: 200

  Routing Bit Set on this LSA
  LS age: 19
  Options: (No TOS-capability, DC, Upward)
  LS Type: Summary Links(Network)
  Link State ID: 0.0.0.0 (summary Network Number)
  Advertising Router: 12.12.12.2
  LS Seq Number: 80000003
  Checksum: 0x8095
  Length: 28
  Network Mask: /0
        TOS: 0  Metric: 1

res

Paul

 

[Rolf Fischer]

Hi Paul,

a very interesting approach, you made me do some reading ;-)

The summary-lsa [<1-16777215>] is an optional part of the max-metric router-lsa command. The latter is the Cisco implementation of RFC 3137's  "Stub Router Advertisement" - another example for confusing naming in OSPF. Here, the term "stub router" is not related to stub areas, but the concept of making a router less preferable for transit traffic within an OSPF domain. So with or without the summary-lsa option, this command will primarily set the metric of all the non-type-3 links in self-originated Router-LSAs to LSInfinity (0xFFFF / 65.535). I just wanted to point out that this will affect all the transit-paths of this router, and we don't know if this is desired here or not.

Regards

Rolf

[paul driver]

Hello Rolf

Fyi -I came across this during my ccie studies and When I labbed  it up it did suggest what your have explained and also due to the fact I was not able to specify a specific peer to apply it towards.

res

paul