EIGRP question

[mmorris11]

In CCO docs much is made of EIGRP's ability to provide "sub-second" convergence via feasable successors. Are feasable successors required at both ends of a route in order for this to work? For example, if router A can get to D through B or C, and C is a feasable successor in A's topology table, then C will immediately be used in A should B fail. My question is: Does D require that C be a feasable successor to A for this to work end to end? If D routes through B and does not have a feasable successor to A then will D not be required to query for a route to A regardless of the fact that A is ready to go?

[Richard Burts]

Mike

EIGRP does not require feasible successor at both ends. Each router makes its own determination of the feasibility test. A can determine that C does qualify as a feasible successor. And that has no impact - and no dependency on whether D considers C as a feasible successor or not.

It is possible that if B fails that A will use the feasible successor and immediately start to route through C while D recognizes the lost route and must initiate a query.

HTH

Rick

[mmorris11]

So if I ping D from A under these conditions, then no echo requests get dropped, but replies will have to wait until D converges a route back to A?

[devang_etcom]

hi...

yes here D must have route A...to send packet back...

regards

Devang

[mmorris11]

So my orginal assertion stands? Bidirectional sub-second convergence requires that a both A and D have feasable successors? The exception obviously being that if D happens to use C as it's only route to begin with. Then there would be no issue. But if A and D both prefer B and only A has C as a feasable successor then the "sub-second" convergence is unidirectional and requires that D also have C as a feasable successor to provide bidirectional "sub-second" convergence.