Redundancy Doesn't Work OSPF/Needs clear ip ospf process

[3Dg4RIP]

Hi, I'm facing an issues in my network. I'm using two types of redistribution: OSPF into OSPF also OSPF into BGP. Everything is working fine while no link failure, so when this is present I lost connectivity with some elements in the network, I need to execute "clear ip ospf process" and the traffic come back. Some guide or sugerence regardind best practice for reditribution also if there some commands to avoid to use the "clear ip ospf process" in order the routers work fine.

The show ip ospf neighbor, show ip ospf database, show ip route seems fine, but the issues when link failure is continue in my network.

Regards

[balaji.bandi]

We need more information about the network and config. Since we do not have visibility how you connected the networks

[ThrowingPackets86]

Agreed with person above, could you pls share ospf topology table and routing table.

Thanks

[Sebastian Fernandez]

Hi @3Dg4RIP 

Since you are doing OSPF to OSPF redistribution, I assume both OSPF process on the same router at certain stage are expected to see redundant prefix and should takeover after an event. But please bare in mind that there is no best route comparison between two distinct OSPF process and that which ever process gets the route first will use it in the routing table irrespective of its cost or type.  And this is something that you can't force over. 

From your story it is clear that after an event, you need to force clear process to make the other OSPF process become valid. You should looking into why that OSPF route is not cleared during the event. There is no clever hack here, unless you want to script the clear process on certain event basis. But you should eye on making network more resilient through proper design.

A topology and configuration on your devices would help in analyzing it further.

Please mark this post helpful if it contributes to your learning.

- 

Sebastian

[3Dg4RIP]

Hi, thanks for your comments. I'm reading some documments where the recommendation is redistribute internal and external Type 1 and Type2, but reviewing in the routers I only have OSPF Internal and in the last AGG Ring External Type 2. Also I found another recommendation where applied a distribution list deny in, in the RR of the AGG Ring, the ACL of this distribution list contain the Lo0 of the RRs.

[3Dg4RIP]

Hi, thanks for your comments. I'm adding a general Topology. We are using Unified MPLS solutions, using differente OSPF process in each Aggregation Ring, also MP-BGP and MPLS LDP. We are redistributing in OSPF the Lo0 of RR, usually in the Cisco reference the every RR redistribute the own Lo0, in our case we redistribute the Lo0 of both RR. Also in BGP we are redistributing the Lo0 of the AGG nodes, and so on. The OSPF in the AGG Ring are working in an open topology, since we are not using an interface in order to close the OSPF area.

Regards