EIGRP - All routes not showing up in show ip eigrp topology all-links

[dk3874]

Everything is working as expected except for one thing. When I do a show ip eigrp topology all-links on R2, specifically looking for a route to 66.66.66.66/32, I am only seeing two routes. One thru R4 and one thru R5, but I'm not seeing a route thru R1.

When I do a show ip eigrp events on R2, I can see poison reverse is preventing the advertisement for 66.66.66.66/32 from R1, but I do not understand why. Making this more difficult to understand is that R2 is advertising 66.66.66.66/32 to R4 and R4 is advertising 66.66.66.66/32 to R2 and both these routes are showing up in both routers from the other. Poison reverse is not being implemented between R2 or R4, so I don't understand why it is being implemented between R1 and R2. 

Also R2 is advertising 66.66.66.66/32 to R1 and R1 is advertising 66.66.66.66/32 to R3, but R3 is also advertising 66.66.66.66/32 to R1 and yet poison reverse is not occurring at all between R1 and R3 for route 66.66.66.66/32.

Why is poison reverse affecting R1's advertisement to R2 for 66.66.66.66/32, but it's not affecting R2 / R3 for route 66.66.66.66/32 and it's not affecting R1 / R3 for 66.66.66.66/32?

Note that the last octet for each neighbor is the same as the hostname number (e.g. for R5, all ip's configured end with .5; for R2, all ip's end with .2)

R1#sh ip eigrp topology all-links | sec 66.66
P 66.66.66.66/32, 1 successors, FD is 156416, serno 94
via 10.10.10.2 (156416/130816), FastEthernet0/0
via 11.11.11.3 (158720/156160), FastEthernet1/0
R1#

Why does R2 only show two routes instead of three. Why is poison reverse in effect here for 66.66.66.66/32, but not between R2/R4 or R1/R3? For R2/R4 and R1/R3, 66.66.66.66/32 is being advertised out the same interface which learned of the route and split horizon is not turned off. Why is this?

R2#sh ip eigrp topology all-links | sec 66.66    
P 66.66.66.66/32, 1 successors, FD is 130816, serno 16
via 13.13.13.5 (130816/128256), GigabitEthernet2/0
via 12.12.12.4 (158720/156160), FastEthernet0/0
R2#

R3#sh ip eigrp topology all-links | sec 66.66
P 66.66.66.66/32, 1 successors, FD is 156160, serno 7
via 14.14.14.5 (156160/128256), FastEthernet1/0
via 11.11.11.1 (158976/156416), FastEthernet0/0
R3#

R4#sh ip eigrp topology all-links | sec 66.66
P 66.66.66.66/32, 1 successors, FD is 156160, serno 5
via 15.15.15.5 (156160/128256), FastEthernet0/0
via 12.12.12.2 (156416/130816), FastEthernet1/0
R4#

[Martin L · Accepted solution]

I think the issue here depends on 3 factors: split horizon (SH), routing by rumor, and Feasibility condition (FC).   

Routing by rumor is one of characteristics of Distance Vector prot.  like EIGRP and RIP.   Definition goes like "Router learns routes from its neighboring routers' perspective and then advertise the routes from its own perspective." So, it is like I make my best routing decision based on the best info I got from you.  Feasibility condition and SH go together sometimes depending on topology.  Making things worst, sometimes output depends on the order in which router/eigrp came up. 

You have 2 triangles (or sort of 2); making R1 on left dependable on what R2 and R5 reporting.  R1 best route is via R2, so SH tells R1 not to advertise that back to R2. In other words, SH depends on current successor route when making decision about networks.  You will not see that route in All-links topology on R2. 

how did u determine that Router is using split horizon?

Re: R4,  I will try to explain more later; 

any possibility of getting gns3 file ? i think just single gns3 2.x project file will include all info including configs. 

Regards, ML
**Please Rate All Helpful Responses **

[Martin L]

Although, I am not 100% certain, will check my notes, doesn't eigrp router advertise only best route?

[dk3874]

I've attached my GNS3 project. 

Martin L. what you are saying makes sense. I looked at each router's best path and it makes sense that the router would not advertise out the interface which it is learning its best path from. So by going thru each router and asking "what is my best path", the output that I am seeing for each router now makes sense. The one thing that I will probably do is add a few more routers to my topology to try and make things a little more complicated and see if this perspective still holds true (I'm thinking that it will).

So R4's best path is thru R5 and of course since loopback 66.66.66.66/32 is directly connected to R5, R5 is not going to accept any other route from any other router for 66.66.66.66/32. R4 though will advertise a route to R2. And R2's best path is also thru R5, so R5 will not accept any advertisement from R2 for 66.66.66.66/32, but R2 will advertise its best route to R4 and R1. R1's best route is thru R2, so R1 will not advertise to R2, but will advertise to R3. R3's best path is thru R5, so again R5 will not accept any route for its own directly connected link, but R3 will advertise to R1.

Previously, I was thinking that R2 should have one successor route and two non successor routes, simply because there were three routers connected to R2, but I can see now why it only has the one successor route and the one non successor route. I need to figure out how to get more routes (by adding new routers or new links) to show up in my all-links topology table to verify that I now have the correct perspective or to further expand my perspective.

I'll wait to hear what you have to say before I mark your answer as the solution.

I tried to upload my GNS3 project, however, I'm getting an error message: "The contents of the attachment doesn't match its file type".

[Martin L]

Sure, but before you create more complex topo, I would change metric wights to include only delay into calculations, no bandwidth.  it should be easier to manipulate metrics and see changes. 

btw. to attach file I think u must zip it first.  my gns3 is broken after update; must set up new version

Regards, ML
**Please Rate All Helpful Responses **

[dk3874]

Thank you Martin L,
Your answer here and your other comments really answered my question and made things clear for me.
I appreciate you taking your time helping me out.

[Martin L]

via 13.13.13.5 (130816/128256), GigabitEthernet2/0
via 12.12.12.4 (158720/156160), FastEthernet0/0

In case of R4 and R2, slit horizon is enabled of course but feasibility condition of EIGRP Dual is what makes decision here. Notice non-successor's RD is larger then successor's metric, feasibility condition fails, route cannot be use as a "back-up".   All-Link topology will show you all routes even those that are neither successors nor feasible successors but number of routes is still limited by SH. So, not all possible routes are shown. To see all, you would have to disable SH on all routers. 

According to some resources EIGRP split horizon is more of optimization improvement rather then loop prevention method like in case of RIP.  If you disable SH, you will have more routes that do not pass FC condition. More routes equals more Dual calculations, more work, more CPU cycles.  I would partially agree; true in case of R4-R2, but in case of left side topology with R1, I would not turn off SH there.  You can lab this w/o SH to see if there are any significant changes. Let us know but don't get "stuck active" on this ;)

Regards, ML
**Please Rate All Helpful Responses **

[paul driver]

Hello

athe reason is most possibly due to the route path via r1 for that prefix isn't applicable for R2 eigrp to state it as a valid feasible successor as R1s advertised distance isn't lower then R2s present successor calculated distance thus is dos t meet eigrp feasibility condition