We have 3 routers RB3011UiAS (v6.35.4) configured with OSPF protocol.
Main router R1 (Core) connected to R2 and R3 Routers. R2 and R3 connected between themselves for backup reason.
Routing for the backup working fine. But in the future we will need to make a backup only for one direction.
We need that if connection to the R2 will be broken - traffic should go to R2 via R3.
BUT when connection to the R3 will be broken - traffic to this router shouldn’t go via R2.
There is parameter in the OSPF Instance called “Redistribute Other OSPF Routes” and it now = “No”. However, the route to the R3 via R2 still exist in the routing table in the Core.
And the traffic still goes via R2 to R3 and via R3 to R2.
Why this route still exist in the core and traffic still goes in both directions?
Redistribute other OSPF means that routes are redistributed from other OSPF instances. In typical scenarios there is only one OSPF instance and “redistribute other OSPF” is not used.
To make one path more specific you need to change interface cost.
Ok. Understood.
Now, how correctly to make our configuration work? May be some recomendation? Because it is not just the backup with OSPF - it should be some one-way backup…
Thank you.
To put it bluntly, your desired configuration is not compatible with OSPF.
OSPF is not a policy-driven routing protocol like BGP. OSPF is an IGP, and the primary objective of an IGP is to maintain full connectivity at all costs, and to do so over the most efficient path available. OSPF works by making a map of your network - it basically makes a list of all routers that are speaking OSPF, and then it makes a list of links between the various routers. Then it has a list of all possible destinations in your topology, and which routers are connected to those destinations. Then for each destination, it finds the best way to go through the links to reach the destination.
OSPF is not a propating list of prefixes that get modified at each router, like BGP. The only thing that’s an exception to this is external routes, which can be filtered to a degree, but external routes have several behaviors about them that are different than standard interior OSPF routes.
Your desired behavior requires two OSPF instances - instance1 should run on routers 1 and 2.
Instance 2 should run on all three routers. Do not redistribute between the OSPF instances.
Link R1<>R2 should be associated with instance1
Links R1<>R3, and R2<>R3 should be in instance2
Notes for Routers 1 and 2:
They should each only have one interface in instance 2 - that being the link to R3.
They will need to place a higher administrative distance on instance 2 - 111 or higher. (probably 111 is the best choice)
They will need to redistribute static and connected routes into ospf2.
All other interfaces should be considered a part of instance1
Router 3 should be a standard OSPF setup.
This is a clunky workaround but it should work with the behavior you desire - I very strongly considered removing it from my reply, as I would never want to run a network this way, but if you insist that this behavior is what you want, then this will work.
There might be a way to do this using OSPF areas instead, and this would at least be more coherent than a split-brain network… If I think of something that should do the trick, I’ll come update the thread.
