OSPF in fully meshed environment

I'm not sure I understand why you do not want those routes advertised. In case a link goes down, would you not want it to route the other way? If there is some bandwidth constraint that I don't know about, you can still raise the metric on certain routes so it will not use them unless it is the only way. If you really want to avoid this, it doesn't matter which area those interfaces are in, and you can just put distribute lists for the routes you don't want advertised.

Hi - Thanks for responding. Yes, I'm familiary with prefix-lists. For security reasons we have other backup links.

Can you help with the more urgent question I have about what area to place the p2p interfaces between areas 1 & 2??

Thanks

I'm not sure I understand your question. They must go in either area

1 or 2, and provided they are not in area 0, it will not cause a complete reconvergence if these links bounced. Flip a coin! Just kidding! Is there something I am missing regarding why it matters which one they are going in? Otherwise, just pick...

if i understand what you trying to do, you are using different areas for the different virtual circuits across the WAN.

simple answer is, dont - put them all in the backbone area.

If you insist in having areas you can use virtual links - but all that means is you "tunnel" the backbone area 0 across the others areas on your WAN - which is not good.

Basically virtual links in your design means it is time to redesign and get rid of them.....

the OSPF topology and "best route" works best within an area - and an area can include dozens of OSPF nodes.

If you must have areas then make each device an ABR and put an area "behind" each one - but unless you have hundreds of routers, or some specific topology reason to need areas then dont use them.

Hi Stephen

Thanks for the advice. I think we're stuck with an area at each site for the moment anyway. Each site's ABR has a p2p interface in the remote area on the other side of the p2p accept for site 2, which I arbitrarily just put the p2p interface in area 2. Every site now is getting routes from the other two sites, except for site 2 which for some reason isn't getting routes from site 1.

Another question regarding the ospf network type. Initially I didn't specify this and all the neighbors are discovered, and, I believe they're all designated as BROADCAST. Is there a downside to this? What is the requirement or advantage of setting them to say, point-to-point non-broadcast?

Thanks

stephen wrote:

I think I missed the crucial point you were making which is that all traffic has to traverse the backbone area, so, I can't send traffic directly from area 1 to 2?

it sort of can - but the spec for OSPF insists that any ABR should be connected to the backbone.

Cisco have written an OSPF extension into some versions of IOS (and wrote an RFC to cover it) that talks about a "fallback" type config where an ABR that loses its backbone link can carry on working.

i think i have seen this happen by accident, but given i dont know which version support it, and what you are running......

"broadcast" means the hello protocol includes finding a neighbour thru multicasts.

Obviously doesnt work if the interface type doesnt believe in multicast.....

The downside is a very slight overhead compared to point to point in setting up the adj.

Not worth the extra config effort IMO to go point to point, unless you need tight control over exactly what is happening.