eBGP eighbors. Do they exchange BGP routes automatically after neighbor establishment?

[danielseijas3010]

I just want to make sure I understand properly, once a BGP neighbor is established, do these neighbors immediately start exchanging routes currently present in their BGP table without any sort of redistribution or "network" commands?

For instance, the ISP has a router with a BGP table full of routes coming from different BGP neighbors. If this router establishes an adjacency with an eBGP neighbor in my local AS that has just been added to the network and has no BGP table yet but it has learned routes using other sources (Connected, Static, OSPF, EIGRP, etc):

1.- Will the ISP router start advertising its BGP routers to the local AS automatically without the need of redistribution or network commands?

2.- Will the router on my AS NOT advertise routes coming from other sources UNLESS I redistribute or apply network commands?

Thanks in advance I hope the question is clear.

Daniel

[Jose Fonseca]

Hi Daniel, 

That's is correct. At the moment you establish the TCP session, whatever the ISP has in their BGP table, that will be advertise to you. Take in consideration that they can apply policy towards your router and that can change depending on requirements. For instance you can ask them to send nothing but the default route, or your internal routes for instance; in case of MPLS connectivity.

In the other hand, if you have 2 BGP neighbors, you will announce the best BGP route you have in your BGP, based on the policies applied per neighbor fashion (in case you want to filter some of the routes to announce). Of course there's a rule for split horizon that need to be respect. 

But if you only have one neighbor in BGP, and you don't have any neighbor statement nor redistribution, nothing will be share with your BGP peer.

Please let me know if you have any further concern.

[Joseph W. Doherty]

(Actually, Jose has answered you questions, but just another answer . . .)

#1 First I assume when your wrote "BGP routers", you meant "BGP routes".  If so, yes and no. eBGP sends its BGP routes to its BGP peers filtered by any outbound policy.  (BGP peers can also filter a peer's received routes.)  It's possible the ISP inserted none, some, or all its BGP routes using some form of redistribution.  It's also possible, the ISP's BGP routes were none, some, or all, obtained from other BGP peers.  So, do you need redistribution or network commands, no, but they may have been used.

#2 By default, again, BGP peers send their BGP routes to their BGP peers, although those routes can be filtered by inbound/outbound policies.  Also again, this is for BGP routes already known by your BGP router, how they become known might have been via redistribution and/or network statements, or they are known to BGP via BGP, i.e. other BGP peers.

BTW, most any ISP will filter out anything you send them across the BGP session, unless they've agreed to accept it, or "know" of it.  For example, they generally won't take any route from you unless its a block they provided you or it's a public block they agreed to accept from you.  (Generally, they only accept routes from you when you're the end destination for it.  If you're another ISP, things get more interesting.)

Basically, BGP is like other routing protocols, by default, it shares its routes with its peers.  Where it differs from usual IGPs, it can have very complex inbound/outbound route filtering (including aggregation of routes), it also can share additional attribute information besides routes, but they have their own set of rules, and you got to work a bit to initially get edge networks into BGP.