OSPF route not replacing RIP route

[daleschultz]

Hi:

I'm not sure what I'm missing with this problem. I have an 877 router with a connection to an ISP providing me OSPF routes and I have a second service provider giving me RIPv2 routes. If I have a route for subnet 10.1.1.0/24 installed in the routing table from the RIP database with a distance of 120 and metric of 4 and I learn of the same subnet from the OSPF table with a distance of 110 metric of 1, why is the OSPF route not replacing the RIP route?

I'm also using redistribution of RIP subnets into OSPF forcing a metric of 115 and OSPF to RIP forcing a metric of 1. This to cause the OSPF to be preferred.

Thanks in advance,

Dale

[srue]

without redistribution between ospf/rip the ospf learned routes will take precedence over the *same* learned routes in RIP. redistribution doesn't affect the local routing table either.

does the 877 have connections to both ISP's directly? or do you have a seperate router for each ISP connection?

[daleschultz]

I figured that is how it should work. I guess the redistribution is influencing how OSPF selects the route to install into the routing table. I'll have to think on that a bit more.

The 877 is being used as a routing protocol converter. The ISP is using MPLS with BGP routing up to the CE router and OSPF to my 877, C router. The 877 is directly connected to the gateway were the RIP routes originate This gateway is actually to a satellite network of remote sites. The ISP on the MPLS side did not want to enable RIP for whatever reason so I had to install the 877 to convert OSPF <-> RIP.

The subnet of interest is 10.1.1.0/24 and it will originate from the ISP MPLS network 99% of the time. But if it does not then it will change and point to the satellite gateway learned via RIP. The transition from OSPF to RIP leaned route works fine, but when the subnet re-appears on the ISP network my 877 is not installing the OSPF route, it sticks to the RIP originated route even though the distance and metrics are less desirable.

Dale

[BrinksArgentina]

I setup a lab but can't reproduce your error. Attached is the config of the 3 routers.

In this lab, if i shutdown 1.1.1.0 on the OSPF link, RIP route was installed and when I restore the link, OSPF replace the RIP route.

Guido.

Please rate all the helpful comments.

[daleschultz]

Yes that's what I expect but I think the redistribution mucks things up.

Here's my relevant config. Let me know (if you try it) what results you see.

Thanks.

[Giuseppe Larosa]

Hello Dale,

the interaction can be more complex of what has been supposed up to now.

When your C877 redistribute the RIP backup route in OSPF this route reaches the MPLS CE router.

On its turn the CE router may have a redistribute of OSPF into BGP so it generates locally a BGP route for the backup route.

When the regular MPLS service is restored the locally generated BGP route has weight 32768 and is preferred over the eBGP route coming from MPLS cloud.

The correct restore of primary route can be achieved by setting a neigh weight of 40000 on the eBGP session on the CE.

if this is done when primary eBGP route is restored it is picked as best in the BGP table.

This triggers the CE router to start to advertise an OSPF external route coming from eBGP and this can be received by C877 that can install it for its lower metric.

Edit:

to verify my guess you can:

you can take

sh ip ospf database external 10.1.1.0

when primary route is up

notice advertising router and route tag

take

sh ip ospf database external 10.1.1.0

when primary route is down

then restore primary route if you don't see the original LSA coming from CE router again the issue is on the CE node

Hope to help

Giuseppe

[rakesh.hegde]

Hi,

Giuseppe is spot on. Just to add to what Giuseppe has aleady said, since locally originated route is the best route , it will be advertised back to your sevice provider PE , potentially causing routing loops.

-Rakesh

[BrinksArgentina]

Dale, as Giuseppe and Rakesh say, the problem is caused by the mutual redistribution between RIP and OSPF.

If you want to solve the problem without involving your service provider, the solution is to maintain two ospf proces on your 877 router and redistribute 10.1.1.0/24 from RIP and from the "point2point ospf process", only to the LAN ospf process.

You can't do this using two areas on the same process because you can't filter type 5(External) LSA's between areas.

Please rate all the helpful comments. Thanks!

[daleschultz]

Hi:

Thanks for your insight. I hadn't thought outside my box and I suspect your explanation is an accurate account of what is happening. When the primary link is restored I don't see any route coming from the CE router. If I clear the route from my table it then shows up, I suspect before the next RIP update arrives.

I'm not sure if I can get the ISP to make a change but I'll ask just the same.

Dale