BGP routes learned but not advertised to other routers in the same AS

[tachyon05]

R1 --- R2 --- R3 --- R4 --- R5

R1 AS 65100 (site A)
R5 AS 65111 (site B)
R2 AS 65000 learned routes from R1 // can PING SVIs on R1
R4 AS 65000 learned routes from R5 // can PING SVIs on R5
R4 AS 65000 learned routes from both R2 and R4 // can PING SVIs on R1 and R

I see R3 is not advertising routes learned from R4/R5 to R2/R1, and R3 is not advertising routes learned from R2/R1 to R4/R5.  This is causing site A can't communicate with site B.  Same is true from site B to site A.

What is needed on R3 in order for it to advertise routes so site A can ping site B and vice versa?

[David Ruess]

Hello,

What you are experiencing is BGPs response to loop prevention, specifically within the AS. Normal loop prevention with eBGP routes (routes learned from another AS #) it would use the AS path. If it sees its AS number in the path it will reject the route. iBGP routes ARE the same AS, so iBGP peers do not send iBGP routes to each other (i.e. learned from same AS). One way to circumvent this you can make R3 a route reflector. If you do that make sure to manipulate the next hop (easiest way is on the neighbor statement) because by default routes learned form an iBGP neighbor do NOT update the next hop and will break routing.

Hope that helps

-David

[paul driver]

Hello
Can you confirm if R2/R3/R4 have IBGP peering to each other, and what IGP is being used to provide that reachability?

Additionally R2/R3/R4 will need to be aware of each others EBGP peers next hop or have those routers advertise themselves to thier ibgp peers as the nexhop for their respective EBGP peers  

R2
neighbor <r2 & r4> next-hop-self
R4
neighbor <r3 & r4> next-hop-self

[MHM Cisco World]

this design is suboptimal, 
the R2 must iBGP to R3 and R4 not only to R3 why because the R3 NEVER advertise the route learn from iBGP to other iBGP peers.

and it also same for R4 to R3 and R2.

the solution is 
1- using IGP between R2-R3-R4 instead of iBGP, redistribute the eBGP into the IGP 
2- using RR in R3 and using next-hop in both R2 and R4 
note:-in point 2 you still need IGP between R2,R3,R4 but without redistribute eBGP into IGP 

[tachyon05]

Thanks Everyone.

Configuring route reflector on R3 helped.  On R4, using SHOW IP BGP, I see routes to subnets on R1 that are learned by R2.  Same on R2 where I see routes to R4 and R5.  However, using the SHOW IP ROUTE on R2 or R4, those same routes to subnets on the other side of R3 are not showing up in the routing table.  What could be the cause of that?  I do not have any prefix list or route maps configured on R2, R3, or R4 to filter routes.  Thanks

[MHM Cisco World]

I ALREADY mentioned note about what do you need if you want R3 to be RR.

Next hop self in R2 and R4 

And need igp between R2'R3'R4

[tachyon05]

Would you please provide an example of how to configure IGP between R2, R3, and R4?
For both R2 and R4, I already have the statement NEIGHBOR [R3] NEXT-HOP-SELF.  The issue is on both R2 and R4, I see routes to the other side of R3 using SHOW IP BGP, but the same routes are not there when using SHOW IP ROUTE.

[tachyon05]

This is resolved after learning NEXT HOP SELF does not work if route reflector is enabled.  Instead, a route map needs to be created and used as follows.

route-map NEXT-HOP-SELF permit 10
set ip next-hop self

router bgp 123
address ipv4
network R2/R4 route-map NEXT-HOP-SELF out

[MHM Cisco World]

this workaround and it work fine 

the trick is next-hop 
the bgp will select best path after check that the next-hop is reachable or not, 
here the next-hop is not reachable that why bgp not select path as best path even if it receive from RR.
the best path show with "">"" in bgp table