OSPF Type 5 LSA Advertisement

[pbelletty]

Hi

I have a scenario where I have three routers on a LAN. Two of the routers (R1 and R2) run OSPF
and have a static route to an external network pointing to the LAN interface of the
third router (R3).

This static route is redistributed in to ospf on both routers (R1 and R2).

When I look in the ospf database I notice there is only one Type 5 LSA for the external route.

If I do a show ip ospf database self only one of the routers is advertising the LSA.

If I remove the redistibute command from the router that advertises the LSA the other router starts
advertising the route.

Any idea why this behaviour occurs. I would have thought that both ASBRs would be advertising the LSA.

Is this something to do with the LAN interface being in OSPF and the LSA forwarding address being non-zero?

Interested to hear your views.

Regards,
Phil

[Giuseppe Larosa · Accepted solution]

Hello Phil,

>>

Is this something to do with the LAN interface being in OSPF and the LSA forwarding address being non-zero?

exactly , load balancing is achieved by having two routes to FA see it a the effective next-hop

as a result of this recursion a single copy of the LSA is enough to support load sharing in the forwarding plane

this behaviour is per OSPFv2 RFC and has been discussed in the past in the forums

it is not cisco specific, juniper routers do the same in same scenario just to say

Hope to help

Giuseppe

[Peter Paluch · Accepted solution]

Hello Phil,

Just to add to Giuseppe's perfect answer, this is indeed related to generation of functionally equivalent LSA-5 which means they describe the same destination, the same metric and the same non-zero forwarding address. In such case, there is no point in having multiple LSA-5 around if they all say the same. RFC 2328 solves this by mandating that if this occurs, only the router with the highest RID will continue to advertise the LSA-5. All other routers will flush their LSA-5 regarding the same destination.

See http://tools.ietf.org/html/rfc2328#page-140 and below - look for "functionally equivalent" phrase.

Best regards,

Peter

[pbelletty]

Folks

Great answers.

Thanks for your help.

Phil