bgp route reflectors and synchronization ????

[gjstem]

Fellow engineers,

I did a lab scenario utilizing bgp router refectors and and ran into a problem in which the router reflector would not reflect routes until I disabled sychronization. Could someone further explain the mechanics behind this.

It doesn't make sense to me how a route reflector passing bgp table routes from one client to another would anything to do with the synchronization between the IGP table and the BGP table????

thanks,

Gregory Stemberger

[yarik_ya]

Syncronization in BGP is useing for waiting BGP to announce the networks untill all routers which are useing IGP will know the way to that networks. So if on all routers in your AS you are useing only BGP and no IGP (IGRP, OSPF, RIP for example), BGP will not announce some network untill all routers will know the route to it via IGP. And because you haven't the IGP, BGP will waite all the time. So you must disable syncronization to give the BGP to announce the network without waiting while the routers in your AS will knew the route to it via IGP.

[sphillip]

Hi Gregory,

the synchronisation rule (my explanation) is that of the route is not learnt via an IGP, then it cannot be installed in the routing table or advertised to it's peers as intermediate devices may not know of the route. This prevents packets being blackholed by a non-BGP speaking router.

Imagine four routers A, B C and D, A and C are iBGP peers, C and D eBGP peers. If A advertises a prefix to C and C advertises this to D, the situation that may arise is if D forwards a packet to the prefix learnt by A, then C can forward the packet to A, but B doesn't know have the prefix, so drops the packet. Hence why synchronisation is used.

HTH - email stevep@cisco.com if you want to discuss further

[ruwhite]

As a general rule, run iBGP peering on the routers which will transit (if you are transiting), and then turn off syncronization.

:-)

Russ