EIGRP 'stuck in active' mode neighbor reset

[yaseenathar1]

My question is that when EIGRP gets to the 'stuck in active' mode and does not receive any response to its query from its neighbors, why does it clear those neighbors and disturb other routing information coming from those neighbors? Shouldn't the neighbor relationship depend on hello packets rather than response to a particular query?

[paul driver]

Hello

My understanding is If a router is in a SIA state then this would indicate a problem somewhere along the eigrp path.
The reason why eigrp queries are made as John states, It’s a way for the router to calculate a better path selection for a lost network.

If a neighbouring router isn’t effected by the originating routers path loss then itself should reply straight away to the eigrp query, However if itself is also affected then it needs to query its neighbours also to calculate a better path and if there is a problem somewhere on the network within this querying time the eigrp timer could expires and SIA will initiate.

I am sure if this is incorrect i will be corrected ( in a nice way , I hope) by my peers on here?

EDIT - Here is a previous post of peters - EIGRP SIA

res

Paul

[Jon Marshall]

Hi Paul

I hope if I ever have corrected you (which I doubt I have needed to) it has always been in a polite way :-)

Jon

[paul driver]

Hello John

I dont mind one bit if i am corrected - Its part of learning, And I see these forums as an extention to my studies.

On csc I do see as one of my peers John, your knowledge seems vast and I have no doubt your experience is the same.

These forums are a great knowledge base for learning and sharing and hope you and others like you are around for some time.

So correct away( if need be) ...lol

res
Paul

[yaseenathar1]

Hello, thank you for your reply. It happens like you said but just as i replied to Jon's answer, if the queries have been forwarded by the neighboring routers and they are still sending hello packets back to the router that initiated the query then what is the need to restart the relationship and risk temporary outage if the routes to other networks are most probably still valid ?

Yaseen

[Jon Marshall]

It's a good question and hopefully Peter (Paluch) may see this and respond with a more detailed answer (or correct any of my mistakes :-))  

As far as I understand it what happens is that when a router makes a route active and sends a query to it's neighbors those neighbors may then need to send the query on to the other routers.

The originating router sets a timer (as do all other routers who send the query) and if this timer is exceeded before it gets a response it drops the adjacency and then restarts it.

The issue is that the originating router has no idea where the problem is if it doesn't get a reply because the router with the problem could be multiple L3 hops away and it has no visibility of this. So from it's perspective if it doesn't get a response from it's neighbor it can only assume there is a problem with that neighbor (even if it is still getting hellos as far as I know) and so it can no longer use any routes that point to that neighbor as the next hop.

In short there is no way for the router to ask it's neighbor if it is the router that is having a problem or whether it is a problem with another router further along the path.

However since version 12.1 of IOS Cisco introduced a feature called Active Process Enhancement that allows the originating router to in effect query it's neighbor.

What happens is that halfway through the SIA timer the originating router sends another SIA request to it's neighbor and that neighbor, assuming it is not the router with the issue,  then responds with an SIA reply informing the originating router that it is still searching for a replacement route.

The originating router now knows that the problem is not with it's neighbor and so does not need to drop and restart the adjacency when the timer expires.

This is only a partial solution to SIAs though as there will eventually be a router connected to a neighbor that is having problems and that router will have to remove all routes via that neighbor and possibly go active on them.

So it can limit the damage done by an SIA but it is not really a solution to the problem and should not be seen as such.

Edit - I should also say I have not seen this enhancement in action as it has been a while since I worked with EIGRP on a production network but that is how I understand it works.

Jon

[yaseenathar1]

Hi, thank you for your reply. In your answer where you say " it can only assume there is a problem with that neighbor (even if it is still getting hellos as far as I know) and so it can no longer use any routes that point to that neighbor as the next hop." If it is still getting hellos from its neighbor, even though it hasn't heard a reply for its query, shouldn't that mean that other routes are still valid (which should be the case) rather than restarting the whole relationship and risking temporary outage? Or would there be any other risk if the relationship is not restarted and other routes are still used?

And thank you for mentioning the 'active process enhancement' it does seem that will reduce the damage to some extent.

Yaseen

[Jon Marshall · Accepted solution]

I understand what you are asking and I can only I guess the logic is that because the neighbor has not responded then it cannot be used for any routes.

Don't forget EIGRP hellos are just used so that each router knows about other routers running EIGRP and can establish adjacencies with them.

It tells the originating router nothing really about the state of EIGRP on the neighbor.

Jon

[yaseenathar1]

Ok, thank you for explaining this behavior.

Yaseen