1 2 3 Previous Next 36 Replies Latest reply: Jun 23, 2013 6:30 PM by Brandon Go to original post RSS
      • 15. Re: EIGRP Split Horizon
        Dinkum

        Hi Keith,

         

        Off topic question here?

         

        I am watching your INE video of BGP and I want initial configs of the class lab you setup so that I can follow whatever you are doing in class. Is it possible that you hand me over the initial config and complete topology diagram.

         

        Many Thanks in advance

        Mahir

        • 16. Re: EIGRP Split Horizon
          Brian

          Aloha Keith,


          Can you explain why the command "no ip split-horizon eigrp x" is not needed in a full mesh when using main physical interfaces, but in a Hub-Spoke topology using main physical interfaces it is required?


          Thanks.

          • 17. Re: EIGRP Split Horizon
            Keith Barker - CCIE RS/Security, CISSP

            Brian wrote:

             

            Aloha Keith,


            Can you explain why the command "no ip split-horizon eigrp x" is not needed in a full mesh when using main physical interfaces, but in a Hub-Spoke topology using main physical interfaces it is required?


            Thanks.

            Hello Brian-

             

            Which routing protocol did you have in mind for this scenario?

             

            If you can provide a simple diagram, the networks involved, as well as the routing protocols used, I would be happy to assist you with the details of why.

             

            Keith

            • 18. Re: EIGRP Split Horizon
              Brian

              Keith,

               

              I believe we are talking about EIGRP here.  I know how to set it up.  I have tested this in GNS3, EIGRP over FR, Hub-Spoke, with a multipoint subinterfaces at the Hub, you must use the "no ip split-horizon eigrp x" command.  I did not think this was needed when the Hub used the main interface.  I could swear I tested this when I posted my GNS3 configs, but apparently I did not with the main interface.

               

              My setup was three routers in a Hub-Spoke topology using the main interface on all three routers and then again with a multipoint subinterface at the Hub as in the example in the new Route book on page 66.

               

              My question is can you explain why the command is not needed when the main interface is used in a full-mesh topolgy, but yet in the Hub-Spoke topology the command is needed.

               

              I have relabbed this up in GNS3 and have verified that YES, you need the command on the HUB router regardless of whether you use the main interface or a multipoint subinterface.

               

              Was just looking for some additional clarification as to why?

               

              Thanks again,

               

              Brian

              • 19. Re: EIGRP Split Horizon
                Keith Barker - CCIE RS/Security, CISSP

                Hello Brian-

                 

                I appreciate the work you have done, thank you.

                 

                If we have a full mesh, then every one is learning the remote networks, directly from their peers on that common network.   With a full mesh we don't have a single router, that learns about a network from site A, via serial interface 0/0, and then needs to send that same network out of the same interface to site B (because with the full mesh, site A and site B are neighbors to each other directly.)

                 

                Again, I do appreciate the work you have done.  You bring a lot of value to this community.

                 

                Sincerely,

                 

                Keith

                • 20. Re: EIGRP Split Horizon
                  Brian

                  Aloha Keith,

                   

                  Thank you for the kinds words.  I really enjoy this community and it is nice to be recognized, specially by someone such as yourself.  Thank you.

                   

                  As for the explaination, thank you, that is what I was looking for.  I too was thinking along these lines.  Let me see if I understand correctly.

                   

                  In the full-mesh topology, eigrp split-horizon is on, but because we have the full-mesh of PVCs on the frame-relay network everyone forms an eigrp neighborship with each other.  So, no need for the eigrp split-horizon to come into play.  In contrast, the Hub-spoke topology, everyone does not become neighbors and therefore, eigrp split-horizon comes into play on the hub router.  Hence we need to issue the "no ip split-horizon eigrp x" to disable this feature to allow the routes to propagate correctly.

                   

                  PS - I have update my Quick Facts document.

                   

                  Thanks again Keith.

                  • 21. Re: EIGRP Split Horizon
                    jamesh

                    I'd like to thank you guys for chewing on this with me. Like I was saying earlier, this collective back and forth does a heck of a lot more for my learning than just reading.

                     

                    Keith and Brian:

                     

                    Really appreciate your input on these forums, your experience helps people like me out tremendously.

                     

                    You guys have a great Christmas and New Year.

                     

                    James

                    • 22. Re: EIGRP Split Horizon
                      Martin

                      See, everyone is confused. This could be one of those Cisco She/He said and/or Unsolved Mysteries. The answer might depend on version of IOS ?

                       

                      Brian Dennis of INE said in one of his CCIE videos (about 17 minutes) there is no magic auto disabling Split Horizon on physical interfaces in EIGRP Hub Spoke topo.

                       

                      Where as, Jeremt Ciora in his CCNP Route video shows us slide (his drawing) with lines: EIGRP Split Horizon is disabled on physical interfaces and enabled on multi-interfaces.

                       

                      mmmm, go figure....

                      • 23. Re: EIGRP Split Horizon
                        Martin

                        Sometimes, DUAL (in debug) will tell you about split horizon as in case of EIGRP from Route video by J.C.

                         

                        DUAL: Doing split horizon on Serial1/0.1

                         

                        DUAL: Dest 10.1.1.2/32 entering active state.
                        *Dec 26 01:28:55.718: DUAL: Set reply-status table. Count is 1.
                        *Dec 26 01:28:55.722: DUAL: Doing split horizon on Serial1/0.1
                        *Dec 26 01:28:55.722: DUAL: Going from state 1 to state 3
                        *Dec 26 01:28:55.726: Packetizing timer expired on FastEthernet0/0
                        *Dec 26 01:28:55.730: Packets pending on FastEthernet0/0

                         

                         

                        Open attached file to see more debuging

                        • 24. Re: EIGRP Split Horizon
                          Keith Barker - CCIE RS/Security, CISSP

                          Martin wrote:

                           

                          See, everyone is confused. This could be one of those Cisco She/He said and/or Unsolved Mysteries. The answer might depend on version of IOS ?

                           

                          Brian Dennis of INE said in one of his CCIE videos (about 17 minutes) there is no magic auto disabling Split Horizon on physical interfaces in EIGRP Hub Spoke topo.

                          Where as, Jeremt Ciora in his CCNP Route video shows us slide (his drawing) with lines: EIGRP Split Horizon is disabled on physical interfaces and enabled on multi-interfaces.

                           

                          mmmm, go figure....

                          Hi Martin-

                           

                          I enjoy your posts.

                           

                          I worked with Jeremy at KnowledgeNet, and I worked with Brian Dennis at INE.   Both are great guys.

                           

                          I wanted to emphasize the importance of labbing up scenarios to reinforce the concepts, regardless of who says what. 

                           

                          Best wishes,

                           

                          Keith

                          • 25. Re: EIGRP Split Horizon
                            Martin

                            continuing my post about various "claims", I found this:

                             

                            ch.7 p.240 in ciscopress CCIE R&S 4th edition:

                            Most interface types enable split horizon by default, with the notable exception of a physical serial interface

                            configured for Frame Relay.


                            and D. Teare writes in Implementing Cisco IP Routing (ROUTE) Foundation Learning Guide (ch.2 after Example 2.27)

                             

                            Split horizon is disabled by default on Frame Relay physical interfaces...

                            Because split horizon is disabled by default on Frame Relay physical interfaces,
                            routes from Router R2 can be sent to Router R3, and vice versa. 

                            Split horizon, which is disabled by default on most interfaces. Split horizon is enabled by

                            default on Frame Relay multipoint interfaces. It can be disabled with the no ip split-horizon eigrp as-number command.

                            • 26. Re: EIGRP Split Horizon
                              Martin

                              Thanks Keith, I also watched your videos for INE and liked those very much

                              (almost as much as those from Scott M, our #1)

                               

                               

                              Point of my labbing is to find out whether you can tell if split horizon is on or off just from running eigrp debugs. "Doing split horizon"message is not bullet-proof. In fact, it does not help you much as I thought; although I read somewhere that debugging could help us.

                               

                              With IP Split horizon On on the WAN (hub), adding interface on Seattle Spoke, does not make or shows debugging output on Denver (other Spoke).

                               

                              WAN1(config-if)#do sh run interface Serial1/0
                                interface Serial1/0
                              ip address 172.19.0.10 255.255.0.0
                              encapsulation frame-relay
                              serial restart-delay 0
                              frame-relay map ip 172.19.0.5 251 broadcast
                              frame-relay map ip 172.19.0.6 250 broadcast
                              end
                              WAN1(config-if)

                              Seattle(config-if)#ip add 192.168.1.10 255.255.255.255
                              Seattle(config-if)#
                              *Dec 26 22:25:54.830: DUAL: dest(192.168.1.10/32) not active
                              *Dec 26 22:25:54.834: DUAL: rcvupdate: 192.168.1.10/32 via Connected metric 128256/0
                              *Dec 26 22:25:54.834: DUAL: Find FS for dest 192.168.1.10/32. FD is 4294967295, RD is 4294967295 found
                              *Dec 26 22:25:54.838: DUAL: RT installed 192.168.1.10/32 via 0.0.0.0
                              *Dec 26 22:25:54.842: DUAL: Send update about 192.168.1.10/32.  Reason: metric chg
                              *Dec 26 22:25:54.842: DUAL: Send update about 192.168.1.10/32.  Reason: new if
                              *Dec 26 22:25:54.846: DUAL: lostroute: do nothing
                              Denver ??? shows nothing; No debug outputs
                              Denver#sh debugging
                              EIGRP:
                                EIGRP FSM Events/Actions debugging is on
                              WAN1(config-if)#
                              *Dec 26 22:25:54.778: DUAL: dest(192.168.1.10/32) not active
                              *Dec 26 22:25:54.782: DUAL: rcvupdate: 192.168.1.10/32 via 172.19.0.6 metric 2297856/128256
                              *Dec 26 22:25:54.786: DUAL: Find FS for dest 192.168.1.10/32. FD is 4294967295, RD is 4294967295 found
                              *Dec 26 22:25:54.790: DUAL: RT installed 192.168.1.10/32 via 172.19.0.6
                              *Dec 26 22:25:54.790: DUAL: Send update about 192.168.1.10/32.  Reason: metric chg
                              *Dec 26 22:25:54.794: DUAL: Send update about 192.168.1.10/32.  Reason: new if
                              WAN1(config-if)#

                               

                              and if I do no ip split horizon on Hub router, Denver shows some debugging info; however, it doesn't straight tell you Not Doing split horizon or something like that.

                               

                              Well, I would not worry about Split Horizon questions over FR on Route exam. I do not think Cisco will put those in because it is not clear whether is on or off as my sources proved.

                              • 27. Re: EIGRP Split Horizon
                                Wesley Kirby

                                Gents,

                                 

                                I would like to throuw a comment/question into this thread if you don't mind.

                                 

                                I am labbing Jeremy C's Route CBT Nuggets from and wanted to play with the Split Horizon theory.  With EIGRP configured and all routers established EIGRP relationship...minus issuing the "no ip split-horizon eigrp xx" command I issued the "debug ip icmp" command on all routers.  I am curious as to why I only see a debug output when I do a successful ping.  For example....when I ping R1 (10.1.1.1) from R3 I see the debug output on both R1 & R3. However, when I try to ping R2 (10.2.1.1) from R3 I do not see any debug output on any router.  I was expecting to at least see a debug output on R1 that the host was unreachable for some reason.

                                 

                                Can anyone explain?

                                 

                                Wes

                                • 28. Re: EIGRP Split Horizon
                                  Martin

                                  R2 and R3 are Spoke routers so they can only communicate via Hub - R1.

                                  First check Layer 2 - Frame relay; then Layer 3

                                  show frame-relay map on spokes show you layer 2 mappings. Do you have mappings to Spokes? Inverse ARP will not help you much; do it manually.

                                  You should be able to ping Spokes' serial interfaces.

                                   

                                  But if you trying to ping Spokes' Loopbacks, then you need to disable Split Horizon.  Spokes do not have route to each other. run show ip route.

                                   

                                  turn on debug ip packets on R3 to see encapsulation error when you try to ping R3 to R2. you do not see any debug info on R1 because packets stops at R3 due to no FR mappings.

                                  • 29. Re: EIGRP Split Horizon
                                    dmorrow

                                    Martin thanks for clarifying this. Having read what you have i.e Diane Teare's most recent CCNP route book suggesting frame-relay enable physical interfaces disable split horizon by default, I was really confused when I labbed a hub spoke topology with eigrp and found that spokes did not see each others subnets. As suggested the "no ip split-horizon eigrp X" fixes that, Accordng to both Diane Teare and Wendell Odom the no ip split-horizon eigrp X only applies to mutlipoint interfaces which have split horizon enable by default.

                                     

                                    I hope you are right and they don't ask this on the exam :-)

                                    But if they do I suspect we would have to answer it as per the book right? even though in practice it doesn't work. Perhaps it is IOS specific, however I tried three different versions on two separate platforms and still no go.

                                     

                                    In the real world I've never deployed this type of configuration it has been generally point to point subifs. So I don't really care about as much but I guess it is frustrating for folks including myself when you are trying to study for an exam and you stumble into these issues, then it's a he said she said situation just as you mentioned earlier.

                                     

                                    It is good to have folks like yourself confirm these issues (at least I know I haven't not gone insane yet).