Routing loops at IPv6 mutual redistributon

[cosmingrosu]

I am doing a lab from 101 Labs book in preparation for ENCOR.
The topology is the following:

and the task requires to configure redistribution on routers R1 and R4, using "the most simple solution that negates the need to filter routes when redistributing".

The solution is from the book is:

The most simple solution that can be used to complete this task is to redistribute the connected RIPng subnet into OSPFv3 and then redistribute OSPFv3 into RIP. This negates the need to implement any filters preventing routes from being re-advertised back to into the same protocol and potentially causing a routing information loop. This task is completed as follows:

R1(config)#ipv6 router rip CCNP
R1(config-rtr)#redistribute connected metric 1
R1(config-rtr)#redistribute ospf 1 metric 1
R1(config-rtr)#exit
R1(config)#ipv6 router ospf 1
R1(config-rtr)#redistribute rip CCNP
R1(config-rtr)#redistribute connected
R1(config-rtr)#exit

R4(config)#ipv6 router rip CCNP
R4(config-rtr)#redistribute connected metric 1
R4(config-rtr)#redistribute ospf 4 metric 1
R4(config-rtr)#exit
R4(config)#ipv6 router ospf 4
R4(config-rtr)#redistribute connected
R4(config-rtr)#redistribute rip CCNP
R4(config-rtr)#exit

I was under the impression that configuring mutual redistribution in multiple points in a network, without employing some kind of filtering (such as filtering based on route tags), will results in network instability and routing loops, since prefixes are re-redistributed from the originating protocol, into the other protocol, and then back into the originating protocol.

Could you please explain me why in this case it doesn't happen? Is there something special about IPv6?

[Harold Ritter]

Hi @cosmingrosu ,

This is actually expected behaviour and it would be exactly the same with IPv4.

The reason the RIP routes redistributed into OSPF are not redistributed back into RIP in this scenario is that OSPF doesn't not redistribute external routes by default. 

The reason the OSPF routes redistributed into RIP are not redistributed back into OSPF in this scenario is that the default administrative distance for RIP (120) is higher than the one for OSPF (110). So since the OSPF routes are preferred and installed in the routing table, RIP can't redistribute these routes back to OSPF.

It is always a good idea though to put filtering in place when doing mutual route redistribution between two routing protocols. 

Regards,

[cosmingrosu]

Thank you for your helping me Harold! I have one more question though:
I spent a couple of hours yesterday looking into redistribution and I identified three core mechanics that redistribution follows:
- for a route to be redistributed from protocol A into protocol B, the route must first be in the RIB installed by protocol A.
- redistribution is not a transitive operation: if redistribution is configured from protocol A to protocol B, and then from protocol B to protocol C, when a route is learned by the router via protocol A and redistributed into B, it will not be automatically redistributed into protocol C.
- a router is able to differentiate between redistributed routes and routes learned via protocols' operations. For instance, if redistribution is configured from protocol A into protocol B, and a route is learned via A which is then redistributed into B, the route that now exists into B's database due to redistribution will not replace the original route learned via A, even in the situations where B has lower administrative distance than A.

My question is: are these conclusions right or am I wrong?

references:
https://www.cs.cmu.edu/~4D/papers/rr-icnp07.pdf
https://netacad.fit.vutbr.cz/wp-content/uploads/ccnp/ce2/ENARSI_Chapter_16.pdf
https://www.cisco.com/c/en/us/support/docs/ip/enhanced-interior-gateway-routing-protocol-eigrp/8606-redist.html

[Harold Ritter · Accepted solution]

Hi @cosmingrosu ,

You are very welcome.

> - for a route to be redistributed from protocol A into protocol B, the route must first be in the RIB installed > by protocol A.

That is correct.

> - redistribution is not a transitive operation: if redistribution is configured from protocol A to protocol B, and > then from protocol B to protocol C, when a route is learned by the router via protocol A and redistributed   > into B, it will not be automatically redistributed into protocol C.

That is also correct. It goes back to the first rule you mentioned. If protocol A has the route installed in the RIB and redistributes it into protocol B, protocol B does not have the route installed in the RIB and therefore does not redistribute into protocol C.

> - a router is able to differentiate between redistributed routes and routes learned via protocols' operations. > For instance, if redistribution is configured from protocol A into protocol B, and a route is learned via A      > which is then redistributed into B, the route that now exists into B's database due to redistribution will not  > replace the original route learned via A, even in the situations where B has lower administrative distance    > than A.

That is also correct. If a route is learnt via RIP, installed in the RIB and then redistributed into OSPF, the OSPF route will not replace the RIP route even though its administrative distance (AD) is lower than the RIP AD, as the router knows that this OSPF route is redistributed from RIP.

> My question is: are these conclusions right or am I wrong?

Correct on all points.

Regards,