1 2 Previous Next 22 Replies Latest reply: Dec 16, 2014 7:55 AM by Shadowjk - CCNP(R&S), CCNA(Security; SP) RSS

    why split horizon needed in EIGRP?

    vipin

      i was thinking on this on my topology, still i couldnt find a need for split horizon in eigrp. Eigrp has a rule that if the reported distance is greater than feasible distance then it wont add the route to topology table. also, if there is an topology change, then it will trigger update to the neighbors. because of this 2 rules, if spilt  horizon is disabled, i dont think there is any problem.

       

      if anyone could explain with an example(with topology in eigrp itself) why split horizon is needed in eigrp would be very helpful.

       

      thanks

      vipin

        • 1. Re: why split horizon needed in EIGRP?
          Uma Shankar V.

          Hi Vipin,

           

          Let me share my views here,please.

           

          Split Horizon required when EIGRP used in frame relay [Hub and spoke] networks.

           

          Consider the topology, as R1,R2,R3 in a network, where R1 is the Hub and R2, R3 are the Spoke routers. So, think what would happen if Split Horizon is enabled on the devices [especially R1].

           

          As R1 is been Split horizon configured by default. When R2 sends an update to R1, it will not send the update to R3, as split horizon is enabled. So, to avoid this Split horizon needs to be disabled on the Hub router.

          • 2. Re: why split horizon needed in EIGRP?
            Elvin Arias

            As you said, the feasibility condition will take care of loops in the network, because of DUAL conservative decisions in regards of the feasibility condition, but one of the reasons why split horizon is enabled on EIGRP is for the query process. If you are routing through a successor, and that successor send you a query for a prefix, considering that split horizon is disabled, you are going to ask right back to your successor about the prefix he is reporting down. This does not make sense and consumes additional bandwidth of the link.

             

            Another case where split horizon is used is to avoid duplicate information inside the topology table. If I send you an update and you send me the same prefixes right back to me, then if you do a show ip eigrp topology all-links you will see duplicate information about the prefixes. This happens on frame-relay point-to-multipoint designs.

             

            Elvin

            • 3. Re: why split horizon needed in EIGRP?
              vipin

              Hi Elvin,

               

              it consumes additional bandwidth(if split horizon is disabled). i understood that.

               

              but i dont think a duplicate information would be there in topology table bcoz it will check reported distance with feasible distance.that one condition will rule out the bounced back prefix to be added in topology table.(i am saying this with my limited knowledge).

               

              vipin

              • 4. Re: why split horizon needed in EIGRP?
                vipin

                Hi Uma gi,

                 

                you replied

                 

                Split Horizon required when EIGRP used in frame relay [Hub and spoke] networks.

                 

                So, to avoid this Split horizon needs to be disabled on the Hub router.

                 

                the above 2 statement looks like contraditing each other, may be i didnt understand what you are saying.

                 

                vipin

                • 5. Re: why split horizon needed in EIGRP?
                  Uma Shankar V.

                  Hi Vipin,

                   

                  Apologies for the misleading words. It should be as Split horizon disabled on physical interface and enabled on sub-interfaces.

                  • 6. Re: why split horizon needed in EIGRP?
                    Elvin Arias

                    The duplicate information would be inside the topology table all links. And this is unnecessary.

                     

                    Elvin

                    • 7. Re: why split horizon needed in EIGRP?
                      Ing_Percy
                      • 8. Re: why split horizon needed in EIGRP?
                        vipin

                        Hi Elvin,

                         

                        r1 is connected to r2.

                        192.168.0.0 network is connected to r1. route is send to r2.

                        when r2 sendsthe route of 192.168.0.0 network back to r1 wont the reported distance of it higher than feasible distance in r1. because of this reason it wont add in topology database.

                         

                        am i right on this???

                        i want to get a clear picture on this. that is why i ask.

                        vipin

                        • 9. Re: why split horizon needed in EIGRP?
                          Elvin Arias

                          As I already said, in the topology table all-links output. See this example as a demonstration.

                           

                          R1 (F0/0:10.12.12.1/24) —— (F0/0:10.12.12.2/24)R2

                           

                          R1 is advertising a loopback to R2.

                           

                          R2#show ip route eigrp

                               1.0.0.0/32 is subnetted, 1 subnets

                          D       1.1.1.1 [90/409600] via 10.12.12.1, 00:03:09, FastEthernet0/0

                           

                          If we check the R1’s topology table, the route would be in R1’s topology table as expected.

                           

                          R1#show ip eigrp topology

                          IP-EIGRP Topology Table for AS(1)/ID(1.1.1.1)

                          Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

                                 r - reply Status, s - sia Status

                           

                          P 1.1.1.1/32, 1 successors, FD is 128256

                                  via Connected, Loopback1

                          P 10.12.12.0/24, 1 successors, FD is 281600

                                  via Connected, FastEthernet0/0

                           

                          But in the case split-horizon is disabled on R2’s F0/0 interface, then R1 will receive the 1.1.1.1/32 prefix from R2, and that is what I said split-horizon prevents on EIGRP. See below.

                           

                          R2(config)#int f0/0

                          R2(config-if)#no ip split-horizon eigrp 1

                           

                          Now, if we verify the topology table of R1, all paths that are feasible would be shown.

                           

                          R1#show ip eigrp topology

                          IP-EIGRP Topology Table for AS(1)/ID(1.1.1.1)

                          Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

                                 r - reply Status, s - sia Status

                           

                          P 1.1.1.1/32, 1 successors, FD is 128256

                                  via Connected, Loopback1

                          P 10.12.12.0/24, 1 successors, FD is 281600

                                  via Connected, FastEthernet0/0

                           

                          This is the same output of earlier, but if we do the all-links, then R1 will be receiving it’s own prefix from R2 as shown below.

                           

                          R1#show ip eigrp topology all-links

                          IP-EIGRP Topology Table for AS(1)/ID(1.1.1.1)

                          Codes: P - Passive, A - Active, U - Update, Q - Query, R - Reply,

                                 r - reply Status, s - sia Status

                           

                          P 1.1.1.1/32, 1 successors, FD is 128256, serno 1

                                  via Connected, Loopback1

                                  via 10.12.12.2 (435200/409600), FastEthernet0/0

                          P 10.12.12.0/24, 1 successors, FD is 281600, serno 2

                                  via Connected, FastEthernet0/0

                                  via 10.12.12.2 (307200/281600), FastEthernet0/0

                           

                          These prefixes are advertised by R2, because they are in the routing table (DV protocols behavior), this is unnecessary and that is one of the reason why EIGRP uses split-horizon. Also other mechanisms are used in conjunction with split-horizon, like poison reverse.

                           

                          Elvin

                          • 10. Re: why split horizon needed in EIGRP?
                            Shadowjk - CCNP(R&S), CCNA(Security; SP)

                            In addition, You are correct in that if split-horizon is disabled the chances of a loop are slim due to the feasibility condition however you need to think about traffic engineering issues. What if someone influences the metric in a complex network? A loop could occur...

                             

                            If EIGRP is confusing then I'd look up RIP loop prevention mechanisms as distance vector loop prevention is the same for RIP and EIGRP. Split-horizion, count to infinity, poison reverse.... etc

                             

                            Josh

                            • 11. Re: why split horizon needed in EIGRP?
                              vipin

                              hi Josh,

                               

                              i was looking for an answer like yours.

                               

                              i cannot imagine a topology where if engineer changes the parameters it would then have duplicate entry in show ip eigrp topology.

                               

                              eigrp is very strong, chances for loop is very very slim.

                              1. reported distance should be less than feasible distance

                              2.it doesnt send perodic updates(OSPF does)

                              3. it will trigger an update when there is a change in the link, but it dont know about the timers for it.

                               

                              because of this 3 rules, chances for loop is very very slim. if loop occurs, it is danger than in RIP.

                               

                               

                              vipin

                              • 12. Re: why split horizon needed in EIGRP?
                                Elvin Arias

                                EIGRP it's a loop free routing protocol. If you want to learn more about EIGRP, read through this document. http://www.cisco.com/c/en/us/support/docs/ip/enhanced-interior-gateway-routing-protocol-eigrp/16406-eigrp-toc.html#splithorizon

                                 

                                Elvin

                                • 13. Re: why split horizon needed in EIGRP?
                                  Shadowjk - CCNP(R&S), CCNA(Security; SP)

                                  Look into redistribution If you change the external AD for EIGRP you could cause both an AD based loop or a metric based loop if the metric specified is incorrectly specified.

                                   

                                  Equally, you can cause a control-plane loop and end up with partial reachability.

                                   

                                  Josh

                                  • 14. Re: why split horizon needed in EIGRP?
                                    Shadowjk - CCNP(R&S), CCNA(Security; SP)

                                    vipin wrote:

                                     

                                    hi Josh,

                                     

                                    i was looking for an answer like yours.

                                     

                                    i cannot imagine a topology where if engineer changes the parameters it would then have duplicate entry in show ip eigrp topology.

                                     

                                    eigrp is very strong, chances for loop is very very slim.

                                    1. reported distance should be less than feasible distance

                                    2.it doesnt send perodic updates(OSPF does)

                                    3. it will trigger an update when there is a change in the link, but it dont know about the timers for it.

                                     

                                    because of this 3 rules, chances for loop is very very slim. if loop occurs, it is danger than in RIP.

                                     

                                     

                                    vipin

                                    The last 2 points wouldn't cause a loop regardless Single area OSPF will never get a loop regardless of the paranoid flooding mechanism.

                                     

                                    Josh

                                    1 2 Previous Next