EIGRP - FC/Split horizon question

mithr

On this topology my question is how are routes for 1.1.1.0/24 on R2,R3 and R4 being created
on second and third occasion

Does it have to do with Feasibility condition or with split horizon rule?

After initial convergence

R4
P 1.1.1.0/24, 2 successors, FD is 158820
via 192.168.24.2 (158820/156260), FastEthernet1/0
via 192.168.34.3 (158820/156260), FastEthernet0/0

R2
P 1.1.1.0/24, 1 successors, FD is 156260
via 192.168.12.1 (156260/128257), FastEthernet0/0

R3
P 1.1.1.0/24, 1 successors, FD is 156260
via 192.168.13.1 (156260/128257), FastEthernet0/0



When i change BW on R4's fa0/0 into 10000

R4
P 1.1.1.0/24, 1 successors, FD is 158820
via 192.168.24.2 (158820/156260), FastEthernet1/0
via 192.168.34.3 (390123/156260), FastEthernet0/0

R2
P 1.1.1.0/24, 1 successors, FD is 156260
via 192.168.12.1 (156260/128257), FastEthernet0/0

R3
P 1.1.1.0/24, 1 successors, FD is 156260, serno 11
via 192.168.13.1 (156260/128257), FastEthernet0/0
via 192.168.34.4 (161380/158820), FastEthernet1/0



when i change also R2's fa0/0 BW into 100

R4
P 1.1.1.0/24, 1 successors, FD is 158820, serno 52
via 192.168.34.3 (390123/156260), FastEthernet0/0

R2
P 1.1.1.0/24, 1 successors, FD is 392683, serno 49
via 192.168.24.4 (392683/390123), FastEthernet1/0
via 192.168.12.1 (25830952/128257), FastEthernet0/0

R3
P 1.1.1.0/24, 1 successors, FD is 156260, serno 11
via 192.168.13.1 (156260/128257), FastEthernet0/0

mithr

i don't believe i understand the point where feasibiblitry condition applies here

for example on second occasion when I change BW on R4's fa0/0 I see a route created on R3 "via 192.168.34.4 (161380/158820), FastEthernet1/0" which is neither a successor nor an FS.

For initial conevergence R4 doesn't advertise back on R3 1.1.1.0 through R2 because of split horizon, right?

Now that I changed BW on R4's fa0/0 why does that? Did R4 forget that one route for 1.1.1.0 had been learned from R3 and sends a second route through R2? Is that the partial update and is it logical to
send it only to R3 and not to R2?

Then I change R2's fa0/0 BW and R4 and R3 lose each one route and R2 learns a new one. Why is that happening?

Maybe I miss some points for EIGRP.Up to now I have read the 3 EIGRP chapters from OLG and I own the FLG too. I believe I understand the FC rule but that BW change totally confuses me.

xXErebuS

mattau wrote: »

yes because the feasiblity condition is that I can use you as a backup route because your advertised distance is lower than my FD
if a links bw is changed you dont care about your metric to the neighbor. You only care about your neighbors FD.

Not sure what he meant by that exactly but your "metric" is your AD which is directly related to your FD. Your changing one of the (if default EIGRP metric calculation is used) K values (bandwidth) that is used to determine the AD and FD which changes your routing table.

Here is a nice calculator that I've used in the past - CalculatorPage6


R3
P 1.1.1.0/24, 1 successors, FD is 156260, serno 11
via 192.168.13.1 (156260/128257), FastEthernet0/0
via 192.168.34.4 (161380/158820), FastEthernet1/0

To start to answer your question did you do sh ip eigrp topology all-links for this or sh ip route?

mithr

I used sh ip eigrp topo all command

xXErebuS

mattau wrote: »

I think you're getting confused with how eigrp does its metric calculation. It only does it on outgoing interfaces.
Also in your examples it looks like you did a show ip eigrp top all links which is another factor.

Think about how eigrp works.
It will send its neighbors its sucessor routes only. and if I learn the route through you I wont send it back to you as per split horizon.
I am not sure what R4 would do with sending R3 its path to 1.1.1.0 if there is 2 successors of equal cost...but for example R4 would send R3 its path to R2 in initial convergence if it was the successor. It doesnt mean that R3 will install the path vis R4, it all depends if it meets the feasibilty condition...its not the end of the world if it doesnt because if R3 to R1 link went down and R3 didnt have a FS path then thats where DUAL would come in to do queries etc.

To not make it too confusing think about what happens when you change R4's fa0/0 bandwidth to 10000.
say 10000 bw = metric 1000 and fast ethernet = 100

R3 will have 1 route in its topology table

Sucessor via R1 FD 200 AD 100

why only 1 route in topology table. Because it only stores sucessor and FS paths.
if you go show ip eigrp top all then it you will see 2 paths via R1 and R4

except just because you change the bandwidth on fa0/0 on R4 doesnt mean that R3 cares that you did. R3 doesnt calculate its metric based on that at all.

for this example R3 will add all outgoing fastethernet interfaces
R3 f1/0, R4 fa1/0, R2 fa0/0, R1 fa0/0
100 + 100 + 100 + 100 = FD of 400 and R4's FD 300

so to recap
R3 sucessor is FD 200/100
R4 400/300

300 which is R4's AD from the perspective of R3 is not less than R3 FD of 200. Hence why it cant be used as a FS because it would cause a loop otherwise.

It can get pretty confusing as its easier if you break it down into simple numbers vs using the large eigrp ones when learning.
Id say just remember that eigrp isnt like ospf, it will only advertise its successor paths, and if I learn a route through you I wont advertise it back to you.

You're giving a lot of miss leading wrong information... EIGRP doesnt "send" route costs for one so how would it know which link is the successor / FS for its neighbor.... ; for two if you do show ip eigrp topology all links which I believe is what he did you will see routes that arent good enough to meet the FS condition (as I had highlighted). If you do show ip eigrp topology you will get the topology that has sucessor and FS....

xXErebuS

when i change also R2's fa0/0 BW into 100

R4
P 1.1.1.0/24, 1 successors, FD is 158820, serno 52
via 192.168.34.3 (390123/156260), FastEthernet0/0

R2
P 1.1.1.0/24, 1 successors, FD is 392683, serno 49
via 192.168.24.4 (392683/390123), FastEthernet1/0
via 192.168.12.1 (25830952/128257), FastEthernet0/0

R3
P 1.1.1.0/24, 1 successors, FD is 156260, serno 11
via 192.168.13.1 (156260/128257), FastEthernet0/0

To answer your question lets look at this one...

When you changed R2 BW to 100 you make the R2 > R1 link have a terrible metric and made it "faster" for R2 to go through R4>R3>R1 instead of directly going to R1. Both routes are there b/c the AD(128257) of the R1 route is not higher than the FD (392683) of the R4 route and its a FS. Split Horizon says R1 can advertise 1.1.1.0/24 network out every link but the 1.1.1.0/24

now b/c split horizon R2 cannot advertise to R4 about 1.1.1.0/24 and R4 cannot advertise to R3 about 1.1.1.0/24.

The reason the AD must be < FD is so there isnt a loop, imagine if R4 had the R4>R2>R1 route as a feasible sucessor; R4 would send to R2 which would send to R4 and so on...

mithr

According to what another user told me, this is happening because of poison reverse

Specifically when a router decides that he has a successor (here for 1.1.1.0/24) he poisons back that route to his successor to eliminate loops.

I found also this link which for EIGRP-split horizon from cisco
Enhanced Interior Gateway Routing Protocol - Cisco Systems

Thank you also guys for helping me