I plan on setting up InterAS option B with ebgp with an external AS and Im hoping someone can help clarify some doubts.
I believe i have a reasonable understanding of how things are working with OptionB, and in my case I think using next hop self to avoid reachability issues will be best since I dont want to advertise the ASBR links IPs into my IGP.
Also this gives me three LSP PE-ASBR-ASBR-PE where the ASBRs will exchange VPN labels via BGP, and the VPN label will be the same within an LSP but swapped to a local label between LSPs.
So far so good, the question is the following:
InterAS Option B states the requirement of directly connected links between ASBRs and eBGP running on that link - eg /30 IPv4, and in XR we then need to add a static /32 host route for the peer IP so that it can be available in the LFIB, as without the static route it wont show up in LFB and therefore break the LSP and the whole thing wont work.
I would however like to run eBGP via loopbacks of the ASBRs instead of on the /30 links IP, since there is more than one link, and this way one eBGP session will suffice and also load share across links. Of course for ASBR1 to be aware of the loopback IP of ASBR2 I will setup static routes for the /32 loopback, with next hop being the link IP, and vice versa.
Note that Im not after Option C here, as Im not using route reflectors.
I would still be exchanging VPN labels between ASBRs via BGP and not LDP, just doing it via the loopbacks.
Would the static route for the loopback satisfy the condition of the nexthop being in LFIB not to break the LSP, even if the link IPs themselves wont have a static route and not be present in LFIB ? or should the links IPs also be in LFIB ?
Since OptionB seems to be quite explicit about a direct link which I still satisfy, but just running bgp as multihop to loopbacks instead of on thee link IPs, will this approach break anything and is it a correct and supported configuration and does it make sense ?
let me know if what im trying to do is not clear and thanks for your help
This will definitely work. You just need to configure one static route to the neighbor loopback address for each and every link you have between the ASBRs. It would looks something like this:
address-family ipv4 unicast
192.168.100.3/32 GigabitEthernet0/0/0/1 192.168.23.3
192.168.100.3/32 GigabitEthernet0/0/0/2 192.168.32.3
thanks for your feedback. The static routes as you mention are exactly what I plan on doing.
in this case Im not expecting that we need the /30 link IPs between ASBRs in the LFIB right ?
documentation for option B never mentions using loopback and egbp multihop so I wanted to reach out to get feedback not to do something that may seem to work but would be an unsupported configuration
im testing this in the lab and will come back with the results
> in this case Im not expecting that we need the /30 link IPs between ASBRs in the LFIB right ?
That is correct.
Do not hesitate to let us know if you have any additional question.
I am sure after lab you will end with you can not do that using loopback and static.
When i do lab the control plane is work label is exchange but when i do ping test the packet is drop at asbr so i search and find the cisco doc. above.
Update me when finish lab.