when these subnets can be propagated by BGP?

[teru-lei]

Hi All,

I practice BGP and I find something strange and I don't know if it's my misunderstand of BGP. The following case:

R1-------------------R2-----------------------R3

lo0 lo0 lo0

R1 is in AS1, R2 in AS2 and R3 in AS3 and they use lo0 as update source. "no syn" command had not used. there is also 1 loopback interface in each router. R1, R2 and R3 are routed by OSPF, and the loopback interfaces and in the OSPF process. Well, now I add a secondary address in loopback interface R1 (1.1.1.1/24)and R3 (3.3.3.3/24) and use "network" command to add the two address to BGP. So in BGP, because the 1.1.1.1/24 and 3.3.3.3/24 is not in the IGP routing table, so R2 should not advertise the two network again. So R1 should not see 3.3.3.3/24 and R3 should not see 1.1.1.1/24. But I find that when I use "show ip bgp" in R1 and R3, I can see both network. Does anybody have idea? To my knowledge, BGP routers will not advertise routes that IGP not know until you use "no syn" command. Am I wrong? Thank You!

Best Regards

Teru Lei

[bbranch]

No for IBGP BGP will propegate the routes regardless of the state of the Syncronisation command, however if syncronisation is enabled BGP will not use the BGP derived route for routing until it has received the route via an IGP. If you look at the routes in the BGP table on the remote router (sh ip bgp 1.1.1.1) you should see it is flagged as not syncronised.

Also if you check the routing table there should be no BGP route in the BGP table.

Your statement would, however, be true for EBGP as it will not advertise routes that are not flagged as Valid and the route will not be classed as valid until an IGP derived route has been received.

[teru-lei]

Hi,

But I really see 1.1.1.1/24 and 3.3.3.3/24 in the routing table with "B" (bgp route). very strange. Also I did not use "confederation" to make the non-optimal routes advertised.

Best Regards

Teru Lei

[thisisshanky · Accepted solution]

Hi Teru,

As bbranch said before, in your case, you are running EBGP between each of the routers. So R2 considers 1.1.1.1 as an EBGP route from R1 and will advertise EBGP route to R3, regardless of whether it has the route in ints IP routing table.

Try changing, R2 and R3 from AS2 and AS3 to AS1. That is put all 3 routers in AS1. Now R3 wont see 1.1.1.1 and R1 wont see 3.3.3.3

Hope that is clear!

[bbranch · Accepted solution]

I did not read your question correctly originally, I now see your peering is all EBGP. Syncronisation does not operate over EBGP bounderies only within IBGP.

An IBGP derived route would only be passed onto an EBGP peer if that route was also learnt via an IGP. In your example you are locally originating the route via a loopback so a route exists in the routing table, therefore this route is passed onto the next AS and then onto the third AS, syncronisation plays no part in this example. Try converting your peering to IBGP, this should work as expected with the route not being entered in the routing table of the remote routers.

[teru-lei]

Hi,

Oh, I see. I understand now. Thank You for your answer and Franky. I met another question of IGP redistribution also and I had post it. Would you please give me some idea?

Best Regards

Teru Lei