Help with OSPF problem

[marcio.tormente]

Hello,

I'm trying to fix a problem with OSPF, but I could find how to solve.

In my topology I have the head office in the 0 area and to talk with the branch (50 area), I have two links, one 10Mb lan-to-lan and other 2Mb serial.

The problem is that, instead of OSPF choose 10Mb link they always choose 2Mb link.

If I put the link between DC-RO-SUC01 and DC-RO-SUC02 ind the 50 area, OSPF works perfect, but I can't left the configurations like this, because I have anothers branchs and all of then make conection with DC-RO-SUC01 and DC-RO-SUC02.

If anyone understood my horrible english and the problem, I'll be very greatful to reveive any sugestion.

The topology is attached.

Thanks

Marcio

[Reza Sharifi]

Marcio,

Which link is 10Mb and which is 2?

does "sh ip os int" show the correct metric for each link?

How are the other branches connect to your core? Are they each in a different area?

HTH

[marcio.tormente]

Hello Reza,

Thanks for your replay.

Router 1 is 2Mb and router 2 10Mb.

Yes, all the metric is correct.

All the others branches are the same, have the same connection, each branch is in a different area.

Tks

Marcio

[ebarticel]

Hi Marcio,

Are the interfaces on branch routers connected to head office in area 0? The branch routers needs to have at least one interface in area 0.

Make sure on the LAN connection that head office router becomes DR, and no BDR(set priority 0 on branch router interface).

Also look at the routing tables on routers and see if any static routes are configured.

I see back up links between your routers, do they have hsrp configured? If yes maybe the router with serial link is the primary.

Hope this helps

Eugen

[marcio.tormente]

Hello Eugen,

Thanks for your support.

All the interface is in area 0, I put one of the in area 0, but all the routes become loop and I lost my access to this router.

I don't have DR election because all the interface is like point-to-point (ip opspf network point-to-point).

I have no static routes, only OSPF and I put the branch as NSSA.

There are some configurations about HSRP, but only in the 6513 not in the router that receive the links.

Tks

Marcio

[ebarticel]

Hi Marcio,

I have another suggestion for you....Why don't you create 2 subnets with 30-bit mask and configured between the branch and Head Office. This will solve the problem with the loops you have. I don't know what speed is your back up links between the routers at branch and the routers ar Head Office.

Second suggestion is create a static route from Head Office to branch on the 10M link. This way you will see if it uses the connection from head office to branch.

The default way OSPF works on LAN is electing DR and BDR. Did you check the neighbor table? It should show there if you have DR or BDR.

Also if you have HSRP configured remember that this is a Layer 2 and the decision is made before L3. If your switch thinks that router connected to serial link is Primary then traffic will go there before OSPF can do anything. Maybe the router uses the switching functions to forward the packet, because of the hsrp settings.

Again the simple way to solve the problem is use 2 new subnets on the 2 connections and assigned the ips to interfaces (subinterfaces) and clear the ospf counters and then you will see that OSPF automatically will add the LAN connection into the routing table.

This will not affect your back up links between routers.

Hope this helps

Eugen

[marcio.tormente]

Hello Eugen,

All the interfaces between all equipaments are using IP address with /30 mask and I put this interfaces as p2p.

I made a LAB with this configuration and I changed the area 50 for area 0 in the connection between Head Office and Branch, and put the the interfaces between both branches int the area 0 too.

In my lab everything works fine, now I have to find out why in the real equipament this kind of configurations makes loop once I did it before.

I'll meke this configuration again and take the logs and I can share here.

Thanks again.

Marcio

[ebarticel]

Hi Marcio,

I have made a drawing ( not very professional) with your 4 routers and I highlighed the areas that I think should be configured to solve the loop problem and have the LAN link as default. You don't need to change any costs. If OSPF is configured all over your network, is a good ideea to have a DR and BDR for each area. It will help with loops problems too, if an interface is flapping (on and off) constantly due to physical problems or misconfigurations.

I hope this it will be of help to you

Have a Happy 2012 by the way

Eugen

[marcio.tormente]

Hello Eugen

In my LAB, the configuration is almost the same like your drawing, the unique difference is that I put the connection between router on the top of topology as area 0 too, only the lan interface (connection with users are in the area 50.

If I left the configuration as you said, one router will send traffic for othe only if the link with H.O fall down.

I hope this week I can do all configuration in the real equipament, because in Lab, everything works fine...

Hapy New year and thaks for your help

Marcio

[ebarticel]

Hi Marcio,

Thats ok, glad to help.

I think the reason you get a loop is that you have area 0 configured between the branch routers. I don't think you want to do load sharring between the ethernet and serial interface.The area 0 configuration on branch routers confuses OSPF and that could explain why you have LAN subnet advertised as O IA, because it is advertised by branch router. Routers within an area have to know about all links within the area. The OSPF gets confused because it receives a route for area 0 coming from router in area 50, and OSPF thinks that is a local link within area, and treats the serial link as the real link between the areas.

All the best

Eugen

[Kishore Chennupati · Accepted solution]

Hi Marcio,

By default ospf would simpley work its best path by SPF algorithm. Looks like somewhere the ospf is not enabled on the interfaces or the metric has been changed. please check "sh ip osp interface summ" on each router and see if the cost and the correct interfaces are in the ospf process.

if you can check that first then maybe we can get a clean idea

HTH

Kishore

[marcio.tormente]

Hello Kishore

Thanks for your support

First os all, I clicked in the wrong botton and your answer was put as the correct, but is not.

The OSPF is working, I can ping all my networks, the problem is only the way that OSPF choose to send information.

Tks

Marcio

[mvsheik123]

Incase all the other BOs have similar config and OSPF working fine, then I would just change the ospf cost on the 2Meg link on both end so that 10Meg link will be the primary.

hth

MS

[marcio.tormente]

Hello mvsheik123

Thanks for your support.

I alredy did this configurations, and the way still remain the same.

By default OSPF choose way first "O" second O IA" . Traffic which going to 2Mb is "O" and to 10Mb is "O IA", this is the reason why nothing change when I put different cost in the interface or routing process.

I believe the Eugen told the correct answer, I just can't understand why the routes make a LOOP.

Tks

Marcio

[Kishore Chennupati]

marcio,

The problem you have is due to a simple reason. What you have is suboptimal routing.

This is what is happening in your case

+++Traffic going from BRANCH to HQ

Lets say RJ-RO-1 is the active hsrp router. However it knows that the best path to the HQ is via RJ-RO-2 because that has the 10M link as ospf would calculate that as you also have a link between them(vlan 100). so traffic would be going the way its supposed to.

+++Traffic going from HQ to BRANCH

This should affect traffic directly connected to the  switches DC-RO-SUC01 and DC-RO-SUC02. what i mean by this is

that DC-RO-SUC01 learns 10.20.240.242(branch prxy for example) via Intra area(O) via RJ-RO-01 and also  Inter area(IA) via CSR-FSP-01. so it chooses the 2M intra area link.

similarly, DC-RO-SUC02 learns 10.20.240.242(branch prxy for example) via Intra  area(O) via RJ-RO-02 and also  Inter area(IA) via CSR-FSP-02. so it  chooses the 10M intra area link.

When you connect the link between the switches and put the link in area 50 then the switch DC-RO-SUC01 now knows

10.20.240.242(branch prxy for example) via the 2M (O) , the link between the swtiches(O) and also the (IA) link as well.

so now OSPF will calculate to see which is the best one. Obviously the it will choose the path that is via the link between the switches and hence it will take the 10M correct path eventually. COST comes into affect when the interfaces are in the same area

HTH

Kishore