OSPF and failure on switched link

Hi,

let’s assume the following theoretical scenario:
ospf.jpg
All routers are connected using /30 subnets, all in the backbone area.
Usually NET1 and NET2 would communicate via the link R1<->R4 as the shortest path.
Now S2 failes, but the physical link is still up on R1 and R4.
What would happen to the network in this case?

Best regards

Max

OSPF wouldn’t be able to receive hellos on the link between R1 and R4 because of the switch being down, and after the dead timer expires (default 40 seconds on broadcast media, can be tweaked) the adjacency between R1 and R4 would be torn down and traffic would go through R1 → R2 → R3 → R4.

That is what I assumed too.
However I had a case in a similar network where R1 and R2 where connected by a switch and NET1 via an additional router on the same switch.
After the failure on S2 the additional router lost all routes that would normally go via R1 and R1 lost all routes (instead of going via R2).

I guess the issue is that both R1 and R4 assume DR for their subnet as the link is still up but they don’t see their neighbor anymore.

No, there’s something else wrong with your configuration or your configuration (possibly subtly) doesn’t match the diagram you posted. This should work out of the box. It’s irrelevant that both assumed DR roles, when there’s no adjacency on the link then they’ll route around it.

In reality the network (the relevant part) looks like this:
ospf2.jpg
The orange lines are wireless links that are inside a bridge interface to prevent OSPF state changes from short disconnects as the recalculation would usually take longer than the disconnect. Those links are the problem.

If wireless on R3 fails it will lose all routes (AFAIR) and R1 will lose all routes that normally go via R3.

I currently don’t have enough routers to reproduce this scenario at home.

Have you solved the problem?
I have the same situation in my net.

No, I haven’t found a “real” solution and I think there is none.

What I did however was set the priority of R2 and R3 to 0 so they would not try to become DR.
That means if they loose connection to the third router in the same subnet they should assume the network is down.

I guess the only other way would be to use a scheduled script to ping some hosts in the network and if that fails set it to down.

So, the problem that you are describing is that when the wireless link on R3 fails, R3 cannot get to the rest of the network, except for R1 and R2?

Assuming that I am describing your problem correctly, it looks like the reason for this is that R3 has an up bridge interface on area0 (backbone) as well as its ethernet interface on the area1 network shared with R1 and R2. When the wireless link on R3 fails, it is essentially splitting your area0 into two … the bridge interface on R3, and then everything else. OSPF does not allow for a packet leaving area0 to enter a sub-area, and then re-enter area0. Try creating an area0 link between R3 and R2, possibly through a vlan through that switch.