I tried several times to recreate a topology that is used by CBT Nuggets in the BSCI BGP series where subnets are redistributed into IBGP. I am able to see the redistributed routes in the BGP table on all IBGP routers and they are all marked as both valid and best routes, but I still do not have end to end connectivity. I also have the BGP router that is advertising the BGP routes into IBGP configured with the next-hop-self command and I am using IOS ver 12.4(13b) so synchronization is disabled by default and I am using ospf as my IGP.
Any help I can get will be greatly appreciated, I do not want to continue until I understand why I am unable to reach the redistributed route.
I have attached the topology diagram.
Thank you all.
Could you provide a specific example for us? In any event we'll need more info, such as the routing table entries for the specific subnet in question, redistribution commands, configurations for both BGP and OSPF.
Are you recreating this in Dynamips/GNS3, real hardware, or a simulator?
I have included a screen shot of R4 & R5 respectively and I am using GNS3. Let me know if there is anything else that's needed.
As soon as the image is released from the queue, I'll take a look.
In the future, hosting it somewhere else on the internet and then linking to it could be more expedient when you're looking for quick responses!
Travis, would mind recommending someplace online where I can host it and then link to it as you suggested?
Some thoughts to try:
1. Ensure there are static routes to get from AS5500 to AS6500 and vice versa
2. Neighbor statements setup correctly, and you should be able to ping each other.
3. If you are using Loopback as their router-id, you must use the command "update source".
4. For IBGP peers to talk to each other, you either **fully mesh them, **use next-hop-self on R4, **use route reflectors, **redistribute BGP into IGP.
Hope this helps.
Except for route reflectors, all else are in use and working perfectly. I do have static routes between the to EBGP routers, with next-hop-self , update-source as the loopback IPs, I have all IBGP routers fully meshed.
Have you tried clear ip bgp * ?
Whats your IGP?
Yes, I have cleared the BGP process several times. My IGP is OSPF.
Thanks for trying to help me with my routing issue. I have included an screenshot of the routing table on R4 and the result of a traceroute from R1 to the redistributed 200.1.x.0 subnet on R5. It appears the issue is on R4, but I have no idea what it could be. Scott you asked about if R5 knew about my routes, correct me if I am wrong. but from what i understood, advertising my internal routes into BGP is not a good thing.
Ok, so what router comes after that 4.2 address? Is that your external peer? Did you look at its routing table to see whether it knew your route back?
If that belongs to an internal router, then your choices are:
1. Redistribute the routes into the IGP. Yup, it kinda sucks, but reachability needs to happen!
2. Have your ASBR give all the internal routers a 0/0 route
3. Tunnel from the router you want to get out to your ASBR router and peer bgp over the tunnel interface.
Reachability throughout the entire path is a must. If someone doesn't know either direction, it's not going to happen!
1- The advertised Networks from R5 doesnt indicate Host address, I mean the access-list conatains Network addresses not the host addresses of the loopback interfaces.
2- R1 wont have the Network advertised by R4 in its BGP table due to the (Split-Horizon) rule. In order for R1 to learn Networks advertised by R5, then a router-reflector or Confederation scenario should be configured to disable the Split-horizon rule.
So, R2 or R3 should be a router-reflector and R1 is the client to reflect the IBGP learned route from R4.
3- You also dont need to set the (next-hop-self) to all neighbors of R4, but rather, you would need the (next-hop-self) to be configured on R3 and R2 towards R4 to set its looback as the next-hop for all learned network from R5.
Check the above points and you should be able to have end to end connectivity.
Couple of things springs to mind:
On R4, sh ip bgp sum ? //are your neighbors in passive or active state?
On R4, sh ip bgp ? could you please check there are bgp routes on R4 AND R5 so the packets can return.
On R1, do the same thing... but check the destination for next-hop should be R4.
If all fails, "be the packet"... Nice Scott. All jokes aside, debug ip packet to see everything on R4