After an eigrp successors metric increase?

EdTheLad

I've read that if the metric of the current successor increases,in order to avoid sub optimal routing, the local router should try and find a better route via diffuse algorithm.This makes sense, but i've noticed that this only occurs on a router that is 2 hops or more from the router originating the network.
I have 3 routers A,B and C, on A i have net 1.1.1.1/32,when i increase the delay from 5000 to 6000, the local router will do a recomputation and place the new FD in the topology and routing tables.
An update with the new distance is sent to B, B receives the new higher distance inserts it in the routing and topology table while maintaining the original FD.
An update with the new distance is sent to C, C receives the new higher distance, see's thats its from its successor(has a hop count >1 ) and starts
a diffuse algoithm.
Is this correct? The only difference i can see between B and C is the hop count! So is the hop count a variable used when deciding to run a diffuse algorithm when the successors metric increases?
Thanks..

Yankee

A recalc is only immediately triggered in EIGRP by a change in BW or delay. Any other changes do not trigger a recalc but will be accounted for during the next recalculation that is done.

Yankee

EdTheLad

Any aspiring ccie student out there know why B doesnt begin a diffuse algorithm and C does? Since an update is received on B with a metric increase for the successor to 1.1.1.1/32, the FD on B will stay the same but the local distance to 1.1.1.1/32 will be increased,since the local distance is greater than the FD and no FSs are present shouldnt B perform a diffuse algorithm? Why B doesnt and C does? Anyone?

Humper

ed_the_lad wrote:

Any aspiring ccie student out there know why B doesnt begin a diffuse algorithm and C does? Since an update is received on B with a metric increase for the successor to 1.1.1.1/32, the FD on B will stay the same but the local distance to 1.1.1.1/32 will be increased,since the local distance is greater than the FD and no FSs are present shouldnt B perform a diffuse algorithm? Why B doesnt and C does? Anyone?

I will try and replicate this tonight. By the way is this using Dynamips or real lab equipment?

EdTheLad

Ok, i found the answer,when the metric is increased on A, since 1.1.1.1/32 is locally connected the FD will be increased and an update sent to B, B will modify its locally calculated distance (FD/AD) pair but leave the actual FD for the route the same(If FD was lower i.e. better it would be modified,since its higher the old FD is kept).If the (AD) stays below the actual FD,B will only send an update as the local calculation still results in a successor.If the value of (AD) raises above the actual FD a diffuse calc is required and queries will be sent.C in this case just happened to have its (AD) exceed the actual FD.
Only took 2 or 3 days to figure out,if i continue like this i should have ccie by 2011 .

Doyles explanation of the diffuse alogorithm is incorrect in both the old and new version.New version diffuse example 2,page 331,332.He ends up with the correct answer but his description of events isnt acurate,this along with other bad descriptions from other sources have caused me great mental pain but now since i understand it, i have a feeling of euphoria!

Webmaster

for finding the answer, and the explanation, which makes sense even to me (the difference in action at B an C that is).

What threw me off in your first reply is: "since the local distance is greater than the FD and no FSs are present shouldnt B perform a diffuse algorithm?"
But there's an FS afterall right (after local comp on ? Which distance is updated, and then B sends updates, but no queries (no diffuse), to all neighbors.

Humper

ed_the_lad wrote:

Ok, i found the answer,when the metric is increased on A, since 1.1.1.1/32 is locally connected the FD will be increased and an update sent to B, B will modify its locally calculated distance (FD/AD) pair but leave the actual FD for the route the same(If FD was lower i.e. better it would be modified,since its higher the old FD is kept).If the (AD) stays below the actual FD,B will only send an update as the local calculation still results in a successor.If the value of (AD) raises above the actual FD a diffuse calc is required and queries will be sent.C in this case just happened to have its (AD) exceed the actual FD.
Only took 2 or 3 days to figure out,if i continue like this i should have ccie by 2011 .

Doyles explanation of the diffuse alogorithm is incorrect in both the old and new version.New version diffuse example 2,page 331,332.He ends up with the correct answer but his description of events isnt acurate,this along with other bad descriptions from other sources have caused me great mental pain but now since i understand it, i have a feeling of euphoria!

That makes sense now....Good work on figuring it out !!!!!! I personally enjoy learning EIGRP!!!!