Hi Adam, The scenario is supported by pmacct, there are two pieces to it:
* pmacct BGP daemon acts as a passive BGP neighbor and replies to an incoming BGP OPEN message with the same AS number contained in the OPEN. This means a single collector can peer with different ASNs no problem. If your NetFlow export model is ingress at edge interfaces facing customers, peers and transits you should be mostly sorted. * If you want to get end to end visibility, ie. a flow from customer in AS111 to a transit in AS222: you collect ingress NetFlow at AS111 but want to see the exit point in AS222, you can use a mix of bgp_follow_nexthop and bgp_agent_map. bgp_follow_nexthop allows to define IP prefixes to be considered "internal" BGP next-hops so granted pmacct peers with all ASBRs, it can follow BGP tables until it hits a "foreign" BGP next-hop; bgp_agent_map is because routers would typically BGP peer with pmacct using one of their loopback interfaces; whereas it is very possible ASBRs of AS111 and AS222 are eBGP peering using their transfer network IP addresses (ie. /30): so bgp_agent_map is useful in this context to map these addresses back to the loopback interface used for the BGP peering. In case something of the above does not work, then it's most probably a bug (or we have to review assumptions) so feel free to follow with me privately for some troubleshooting. Cheers, Paolo On Mon, Nov 18, 2013 at 12:53:41PM +0100, Adam Bogdan wrote: > Hi > > I have small problem with pmacct implementation > > I have network with 3 ASes - in each AS there is at least 2 routers, > sometimes more > Each of these ASes hold some part of the full BGP table > > It looks like this: > AS 111 - R1,R2 > AS 222 - R3,R4 > AS 333 - R5,R6 > > R1 is connected to uplink1 and uplink2 > R2 is connected to uplink3 and uplink4 > R3 is connected to uplink5 and uplink6 > and so on > > Routers in ASes are connected via ibgp and each router is connected with > each other via ebgp > > If I connect customer to any of these ASes he's able to receive all bgp > routes or some part of it - depends to which AS I'll connect him. > > And now I'd like to apply pmacct to see how traffic is flowing between > uplinks and customers, which AS to which AS is generate > what amount of traffic. > > Question is how to do it ? If I set a machine with quagga I can only > configure 1 BGP session - should I set 3 collectors, one per AS. > But then could a problem with peer_ip_src or peer_ip_dst. Any ideas ? > > Regards > Adam > _______________________________________________ > pmacct-discussion mailing list > http://www.pmacct.net/#mailinglists _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists
