From: MOLINA, MARTIN J *Internet* (PBI) (mm1343@xxxxxxx)
Date: Mon Apr 01 2002 - 17:04:58 GMT-3
In this scenario the use of a route-reflector was not mandatory so I
basically corrected the problem by configuring a full mesh BGP topology
without the reflector. I'll check the archives. I thank everyone for the
input.
-----Original Message-----
From: Gregg Malcolm [mailto:greggm@sbcglobal.net]
Sent: Monday, April 01, 2002 11:31 AM
To: ccielab@groupstudy.com
Subject: Re: BGP SYNC/RR with OSPF (Very Long)
As others have suggested, a confed is a potential solution to the OSPF/BGP
RID issue. My other guess was to config RR clusters so that R1 will learn
your loop from R3 via BGP directly. I haven't tried it but I'll give it a
shot later on just as a learning experience.
Can you please tell us the constraints of the problem ? Do you have to use
RR's and no confeds ?
There are many post regarding this situation in the archives. I did my
search on "BGP RID". Here's one :
Subject: Re: BGP Synchronization and OSPF IGP
From: "Tim Wilhoit" <tilimil@xxxxxxxxxxx>
Date: Mon, 21 Jan 2002 00:22:27 -0600
References: <399E34A117EED311A2FF009027404A6F51D037@xatl08.atl.hp.com>
----------------------------------------------------------------------------
----In this scenario, the best option may be to turn your as into a confederation instead of using route reflectors and have every router be a sub AS and you wont have the sync problem. Otherwise you have to make some careful placement of your EBGP peers into the AS? Any other ways you can get around this problem?
----- Original Message ----- From: "DOLLING,ERNESTO (HP-Argentina,ex1)" <ernesto_dolling@hp.com> To: "'Adam Quiggle '" <aquiggle@nc.rr.com>; "'Peter van Oene '" <pvo@usermail.com>; "'Andy XueWen Qin '" <aqin@cisco.com>; <ccielab@groupstudy.com> Sent: Sunday, January 20, 2002 6:43 AM Subject: RE: BGP Synchronization and OSPF IGP
> People, > > I totally agree that a route inyected into OSPF (typically E2 type) would > have a router-id origin of the ASBR who did the job, and will not change in > the entire OSPF domain. > So far, so good, but what about the router-id of the BGP peer advertising > the prefix??? > > In a fully mesh iBGP environmet no problems will occur, since every BGP > speaker will obtain the prefix from the same ASBR router, and the only > concern is to be sure "BGP router id" and "ospf router id" match in the > ASBR. > > But what about not iBGP fully meshed peers (like route-reflector > environments??) Here a problem could arise. Let's supouse the ASBR is a > spoke (route-reflector client) in the route-reflector cloud. Other > route-reflector clients will receive the prefix trough the > "route-reflector", and will not sync, since OSPF has the "router-id" from > the ASBR. > > Please let me know if I4m wrong here. > > Ernesto. > > > -----Original Message----- > From: Adam Quiggle > To: Peter van Oene; Andy XueWen Qin; ccielab@groupstudy.com > Sent: 1/20/02 1:28 AM > Subject: Re: BGP Synchronization and OSPF IGP > > Peter, > > Because not all the routers in your AS are running BGP and you are > a transit AS. > > AQ > > At 11:23 PM 1/19/02, Peter van Oene wrote: > >Hey Adam, > > > >Why would you learn routes via EBGP that already exist in your IGP? > > > >At 01:57 PM 1/19/2002 -0500, Adam Quiggle wrote: > >>Andy, > >> > >>I believe you are right and that my analysis is flawed. What if > >>the scenario were: > >> > >>R1---(eBGP)---R2---(Area0)---R3---(Area1)---R4 > >> > >>and R4 were the source for the BGP route and an internal OSPF > >>route then I think that situation described below would apply. > >>Do you agree? > >> > >>AQ > >> > >>At 01:49 PM 1/19/02, Andy XueWen Qin wrote: > >>>See my comments in line. > >>> > >>>Thanks > >>>Andy > >>> > >>>On Sat, 19 Jan 2002, Adam Quiggle wrote: > >>> > >>> > Date: Sat, 19 Jan 2002 11:32:17 -0500 > >>> > From: Adam Quiggle <aquiggle@nc.rr.com> > >>> > To: ccielab@groupstudy.com > >>> > Subject: BGP Synchronization and OSPF IGP > >>> > > >>> > Ok gang, thanks to the postings of EA Louie on this subject and > >>> > some lab work I believe that I have an understanding of how to > >>> > make this work and the gotchas. Let me see if I can summarize, > >>> > and as always if I'm wrong...somebody let me know. :-) > >>> > > >>> > Scenario: > >>> > > >>> > R1---(eBGP)---R2---(Area0)---R3---(Area1)---R4---(eBGP)---R5 > >>> > > >>> > BGP > >>> > ----------------- > >>> > AS1 - R1 (RID - 1.1.1.1) > >>> > AS2 - R2 (RID - 2.2.2.2) > >>> > AS2 - R3 (RID - 3.3.3.3) > >>> > AS2 - R4 (RID - 4.4.4.4) > >>> > AS3 - R5 (RID - 5.5.5.5) > >>> > > >>> > OSPF > >>> > ----------------- > >>> > R2 is an ABR and ASBR (RID 2.2.2.2) > >>> > R3 is an ABR and (RID 3.3.3.3) > >>> > R4 is an ASBR (RID 4.4.4.4) > >>> > > >>> > Stipulation: Inject BGP routes learned from R5 into OSPF and > >>> > dynamically propagate them to R1. You can't use "no sync" on > >>> > R2. > >>> > > >>> > Problem: Because you must use synchronization on R2 the BGP RID > >>> > and OSPF RID must match before a route will get advertised to R1. > >>> > However, because R2 and R4 are in different OSPF areas, R2 will > >>> > never see a route sourced from R4, it will always see the routes > >>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > >>> Are you sure ? If you redistribute the external > >>> routes on R4 into OSPF, it'll always show up as > >>> originated from R4 in the whole OSPF domain. > >>> R3 is just playing a role of passing LSAs, it won't > >>> change the route's originator to himself. > >>> I tested it, the BGP/OSPF ID confliction won't be > >>> a problem in your scenario. > >>>----------------------------------- > >>> > being sourced from R3. Thus, even if the BGP routerid and OSPF > >>> > routerid on R4 are identical, the routes will never get sent to > R1. > >>> > > >>> > Solution: Change the OSPF RID on R3 to 4.4.4.4 and OSPF RID on R4 > >>> > to something else such as 9.9.9.9 (so that you don't get > conflicting > >>> > OSPF RIDs). Now R2 will see the BGP RID of a route from R4 as > 4.4.4.4 > >>> > and learn the OSPF route from RID 4.4.4.4 and the route will get > >>> > sync'd and subsequently sent to R1. > >>> > > >>> > Now, I know this works, but I'm not sure if the rationale is > right. > >>> > Subsequently I must now ask what about the other various IGPs that > >>> > could be used. Seeing how RIDs don't play a big part in RIP, IGRP > >>> > and EIGRP I don't see the same "gotcha" from these IGPs. > >>> > > >>> > Comments? > >>> > > >>> > Thanks, > >>> > AQ
----- Original Message ----- From: "MOLINA, MARTIN J *Internet* (PBI)" <mm1343@pbi.net> To: <ccielab@groupstudy.com> Sent: Monday, April 01, 2002 10:44 AM Subject: BGP SYNC/RR with OSPF
> Group, > I understand this has been hit hard lately but I am still not getting it to > work: > > My topology is simple: > R1-------------R2-----------------R3 > > I am using one AS and I am using R2 as a RR. My IGP is OSPF. R1 and R3 have > loopback IP's that have been installed in BGP via the network command. I am > redistributing BGP into OSPF on R1 and R3 so OSPF is aware of the loopback > IP's. The key is I am trying to do this without using the "no sync" command. > I can't get R3's BGP table to sync on R1's loopback network and I can't get > R1's BGP table to sync on R3's loopback network. I have made sure that the > BGP and OSPF RID's are the same on all three routers. The IGP table is aware > of all routes with the proper masks. I changed my IGP to Eigrp and it worked > like a champ so I know it is an OSPF related issue. Does anyone know what I > might be doing wrong? Any help would be appreciated. > > Martin Molina > Senior Network Engineer > Pacific Bell Internet Services > CCNP CCDP > desk: 925 973-7774
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:57:51 GMT-3