OSPF routing problem

GT-Rob

Just thought I would throw this out there, as I am not getting far with it.


I have two WAN links that I can get to my branch sites with. I want to treat these as equal in my internal network, so that I can route traffic over both, so I want both routes installed. The branch routes are learned by my WAN routers via BGP, and redistributed into OSPF, which is what the internal LAN uses. So, OSPF makes the routing decisions.


Now, my internal router is seeing both routes in the OSPF database. Lets use the network 10.0.0.1 for example. However, OSPF is only installing the 1 route, out one of the links (leaving the other one idle).

Metric is 1 on both, and they are redist'd as E2 routes, so cost shouldn't be an issue. There is no other funny business with route maps or editing interface costs or anything like that.


I will note, that the two WAN routers are on different networks (vlan 5 and 6 lets say). However, everyone is in area 0, and like I said the core network is seeing LSAs for both networks, but only installing the 1. A show ip ospf data external 10.0.0.1 shows the two routes with nothing different between them.


Also, the max paths is set to like 6, so thats not the issue.


Any ideas what I am missing here?

networker050184

If you are redistributing from BGP into OSPF, OSPF will only take whats in the routing table. I'm guessing BGP is only putting the one path there so thats all that OPSF is pulling in from BGP. Thats the first thing that popped into my mind.

EDIT: Maybe I'm thinking about this wrong. Can we get a diagram?

GT-Rob

Well the core routers are not running BGP, they are only learning OSPF. And the routers are seeing both routes in the database, I just don't understand what its using to 'choose' with route goes in the RIB.



route A -- WAN1 --- CORE --- WAN2 -- route B


Where route to WAN is BGP, and WAN to core is OSPF. However route A and B are actually the same subnet, just different paths of reaching it.

ilcram19-2

GT-Rob wrote: »

Well the core routers are not running BGP, they are only learning OSPF. And the routers are seeing both routes in the database, I just don't understand what its using to 'choose' with route goes in the RIB.



route A -- WAN1 --- CORE --- WAN2 -- route B


Where route to WAN is BGP, and WAN to core is OSPF. However route A and B are actually the same subnet, just different paths of reaching it.

try to create a route map and an acl to remove the route that you dont want on OSPF see if that fixes the problem

GT-Rob

Well I don't want to block the route. Route A and B are technically the same destination, just with different next hops. The OSPF database sees both, but for some reason picks route B for the RIB. If I shut down or block the routes from WAN 2, then the routes from WAN 1 jump in no problem. I feel like there is some kind of metric or something I am forgetting.


AD is 110 on both. Both are E2 routes.

keenon

just thinking off the fly.. both routes are in the database and are redist as E2.. if you remove 1 and the other pops in.. my first thought is metric

diagram would be greatly helpful

networker050184

I did a basic lab off your set up and both routes are getting in the table on my "CORE" router. I guess some more info like configs would be needed to know more whats going on. Like keenon said, if the other takes over fine, metric is my thinking as well.

GT-Rob

Which makes 100% sense, since theres no other reason for it not to get installed in the RIB, the other route must be a worse metric or AD. But they both show up good.


Im going to do some clean up changes this weekend (hardcoding router-ids, ospf priorities, etc) and maybe do a restart on the OSPF processes, and see how things look.

GT-Rob

Ok, so here is the output on one of our core routers.

#show ip ospf database external 10.192.60.0

Routing Bit Set on this LSA
LS age: 1615
Options: (No TOS-capability, DC)
LS Type: AS External Link
Link State ID: 10.192.60.0 (External Network Number )
Advertising Router: 10.250.7.2
LS Seq Number: 80000050
Checksum: 0x80F1
Length: 36
Network Mask: /24
Metric Type: 2 (Larger than any link state path)
TOS: 0
Metric: 1
Forward Address: 0.0.0.0
External Route Tag: 65500

Routing Bit Set on this LSA
LS age: 943
Options: (No TOS-capability, DC)
LS Type: AS External Link
Link State ID: 10.192.60.0 (External Network Number )
Advertising Router: 10.250.18.20
LS Seq Number: 8000000C
Checksum: 0x4F4A
Length: 36
Network Mask: /24
Metric Type: 2 (Larger than any link state path)
TOS: 0
Metric: 1
Forward Address: 0.0.0.0
External Route Tag: 65500


#show ip route 10.192.60.0
Routing entry for 10.192.60.0/24
Known via "ospf 100", distance 110, metric 1
Tag XXXX, type extern 2, forward metric 1
Last update from 10.250.18.20 on Vlan53, 06:26:10 ago
Routing Descriptor Blocks:
* 10.250.18.20, from 10.250.18.20, 06:26:10 ago, via Vlan53
Route metric is 1, traffic share count is 1
Route tag XXXX



What I want to see in the sh ip route, is both 10.250.18.20, and 10.250.7.4 (which you can see in the database).


PS the OSPF config is very basic.

billscott92787

Right which you should be able to see since the cost of the routes are the same, as well as the fact that the metric types are both type 2 external routes. So, "technically" it should install both routes into the routing table and load balance traffic. Which I'm assuming is what you are trying to accomplish. So when you did the redistribution you set their metrics both to 1. Do you have any other information you can provide, like configs of the redistribution or anything else?

jason_lunde

I think you are running into the "forwarding address" problem...never ran into it myself, but have heard other people complain of it. Heres a cisco link:
Common Routing Problem with OSPF Forwarding Address [IP Routing] - Cisco Systems

What does your "show ip ospf border-routers" on the core show?

HTH's

networker050184

Nice link jason_lunde, I'll have to give that a read.

As far as your OSPF database it looks pretty much the same as the one I have in my lab. I do notice you don't have router IDs set to loopbacks which might have something to do with that link jason_lunde posted.

R5#sho ip o d ext 10.13.13.0

            OSPF Router with ID (5.5.5.5) (Process ID 1)

                Type-5 AS External Link States

  Routing Bit Set on this LSA
  LS age: 1281
  Options: (No TOS-capability, DC)
  LS Type: AS External Link
  Link State ID: 10.13.13.0 (External Network Number )
  Advertising Router: 2.2.2.2
  LS Seq Number: 80000005
  Checksum: 0x483A
  Length: 36
  Network Mask: /24
        Metric Type: 2 (Larger than any link state path)
        TOS: 0
        Metric: 1
        Forward Address: 0.0.0.0
        External Route Tag: 1

  Routing Bit Set on this LSA
  LS age: 1417
  Options: (No TOS-capability, DC)
  LS Type: AS External Link
  Link State ID: 10.13.13.0 (External Network Number )
  Advertising Router: 4.4.4.4
  LS Seq Number: 80000003
  Checksum: 0x106C
  Length: 36
  Network Mask: /24
        Metric Type: 2 (Larger than any link state path)
        TOS: 0
        Metric: 1
        Forward Address: 0.0.0.0
        External Route Tag: 1

R5# sho ip ro 10.13.13.0
Routing entry for 10.13.13.0/24
  Known via "ospf 1", distance 110, metric 1
  Tag 1, type extern 2, forward metric 64
  Last update from 45.45.45.4 on Serial1/0.54, 00:05:47 ago
  Routing Descriptor Blocks:
    45.45.45.4, from 4.4.4.4, 00:05:47 ago, via Serial1/0.54
      Route metric is 1, traffic share count is 1
      Route tag 1
  * 25.25.25.2, from 2.2.2.2, 00:05:47 ago, via Serial1/0.52
      Route metric is 1, traffic share count is 1
      Route tag 1

GT-Rob

Thanks for the link Jason. I will have to read it in more detail when the beer isn't flowing (team Canada plays in an hour, just getting 'ready' lol).

On a show ip ospf border, the core shows both ASBRs as intra-area routers, however the [number] on the route not working is much higher... Not sure what that number means though.


Now I am thinking that OSPF thinks there is a loop. Technically there is, since I want to use two ASBRs to reach the same network.


EDIT the number is the COST!!! So for some reason my routers are seeing the 'working' route as a cost of 2, and the 'non-working' route, as a COST of 33!! I think if I fix that (make them equal), we will be back in business!

jason_lunde

GT-Rob wrote: »

EDIT the number is the COST!!! So for some reason my routers are seeing the 'working' route as a cost of 2, and the 'non-working' route, as a COST of 33!! I think if I fix that (make them equal), we will be back in business!

Thats what I was lookin for man! Congrats, hope that solves it for ya. Congrats on the 500m. speed skating victory by the way.

GT-Rob

So adding a cost of 32 to one side did make it balance out a bit, but I am still not sure WHY the one route had an extra cost. 32 would make it seem like it crossed some 10mb links (default cost calculation), but its 1000/100mb the whole way. I have some new devices in the mix, is it possible they updated the auto-cost calculation with a recent IOS? maybe based on 1gig (but again, only is effected one way, the other was is a cost of 2!).


Anything other than speed (and static ip cost assignments) that would effect the cost like that? Theres only place in my whole network that is 10mb, and its our dialup VPN (not even close to included in these routes).