Hi,
R1 has little choice - it has to go through R2. However, R2 does have a choice, and I personally believe that R2 will use both paths through R3 and R4, assuming the link costs are equal.
R2 will see a single LSA5 originated by R5 and flooded through Area 0. While R5 will also originate an LSA7 into the NSSA area, this LSA will have the P-bit cleared as required by RFC 3101 (RFC 1587), and so will not be translated into LSA5 by R3. We should therefore focus on the way R2 will handle this single LSA5.
R2 will see two LSA5, one received through Area 0 originated by R5, the other originated by R3 as a translation of LSA7 into LSA5. The LSA5 from R5 has its Forwarding Address set to 0.0.0.0; the LSA5 from R3 has the forwarding address set to R5.
On R2, the process will be:
- SPF will start processing the LSA5. Because this LSA5 has the Forwarding Address set to 0.0.0.0, SPF will be looking for paths to R5 which have to be either intra-area or inter-area OSPF routes.
- RFC2328 Section 16.4 Step 3, together with Section 16.4.1, states that to reach an ASBR, we first prefer intra-area paths through non-backbone areas (but we have no such paths), and if there are none, we consider any other intra-area backbone or inter-area path. This has left us with the intra-area backbone path R2-R4-R5, and the inter-area path R2-R3-R5 which are both of the same cost. Section 16.4 Step 3 further says that in case of multiple least cost paths, we should choose the one with the highest associated area ID. However, for R2, both paths go through Area 0, so we're left with both routes.
- Section 16.4 Step 5 says that if the destination 5.5.5.5/32 is not yet in the routing table, install it there with the next hop set to the list of next hops we have identified for the path to the ASBR, and in our case, it should be both R3 and R4. Here, processing of the LSA5 originated by R5 should be finished. Note that by this step, we already have two paths to 5.5.5.5/32 installed in the routing table.
So, in effect, if this analysis is correct, you are using both paths to reach 5.5.5.5/32, assuming that the link costs are equal.
Would this align with your observation if you configured this lab?
Best regards,
Peter
EDIT: Corrected a misleading explanation that assumed that LSA7 originated by R5 will be translated on R3 - it won't because R5 is an ABR and originates a LSA5 for 5.5.5.5/32 into the backbone already, so there is no point in having the same LSA7 translated into another LSA5 by R3. That's why the P-bit in LSA7 will be cleared, preventing it from being translated into LSA5.
