Solved: ebgp-multihop

wgranada1 · ‎05-11-2008

I understand that when you use ebgp-multihop you need use static routes as well. Maybe I'm not understanding this or not doing it right but when I put in the static route I can ping the distant end of 10.128.184.105 but when I look to see if bgp establish it isn't. I see it active but the neighbors never establish. What am I missing?

cisco_lad2004 · ‎05-11-2008

You just need to have a route to your peering neighbor address. For EBGP, typically it is another AS, therefore no IGP , so static routing is the solution.

when BGP is active, it means have not reached established state. to troubleshoot, you need to veriffy the following:

1-can u ping 10.128.184.105 with an extended option sourced from Loopback0.

2-do u have any ACLs blocking TCP sessions.

3-make sure AS64637 are also using ebgp multihop.

4-check if AS64637 are using BGP authentication.

5-check ur logs for any additional info.

Please rate all helpful posts.

Sam

View solution in original post

lamav · ‎05-12-2008

Its OK, dont be sorry, Just be grateful by giving us 5 points each for our posts. :-)

VL

View solution in original post

cisco_lad2004 · ‎05-11-2008

You just need to have a route to your peering neighbor address. For EBGP, typically it is another AS, therefore no IGP , so static routing is the solution.

when BGP is active, it means have not reached established state. to troubleshoot, you need to veriffy the following:

1-can u ping 10.128.184.105 with an extended option sourced from Loopback0.

2-do u have any ACLs blocking TCP sessions.

3-make sure AS64637 are also using ebgp multihop.

4-check if AS64637 are using BGP authentication.

5-check ur logs for any additional info.

Please rate all helpful posts.

Sam

lamav · ‎05-11-2008

Warren:

This post is a bit long but necessary. Kindly read it carefully to its end.

I saw your post and thread yesterday and I saw the confusion you are dealing with.

You talk of static routes and then you get the applications and requirements confused. Let's discuss the application that you are struggling with. We can discuss the other application of static routes in BGP later in this thread, if you like.

First, the situation you are talking about is one in which the BGP neighbors are not directly connected. Remember, eBGP requires that the neighbor addresses sit on the SAME data link - same subnet.

However, when you use a loopback interface on each router to establish a BGP peer relationship, you have some extra configuring to do because they do NOT sit on the same subnet. Underdstand?

The 2 routers may have their WAN interfaces sitting on the same subnet, BUT you are doing the peering using the loopback interfaces, NOT the WAN interfaces. So, when we say the peers are not on the same data link, and we therefore have to use the eBGP multihop solution, we are specifically talking about the loopback interfaces that are being cited in the BGP neighbor statements. Get it?

OK, so why would you not want to peer using the physical WAN interfaces if they are already on the same subnet? It seems counter intuitive to use the loopback interface. Well, the most prevalent reason is to achieve load balncing across multipe WAN links between the two BGP-speaking routers.

Why cant you accomplish load balcning with just setting up 2 BGP peers across the WAN links, one for each link, and be done with it? Answer: Remember, BGP will select the BEST path between 2 routers for any particular prefix. So, even if multiple WAN links exist between the two routers, only ONE BGP path will be selected as the best path and it is on this best path that traffic will flow. So there goes the benefit of that second link!

OK, so the solution is to confuse and fool your router. Set up ONE BGP peer using the loopback interfaces, exchange BGP information, and use the 2 WAN links to actually send the traffic to the other side.

But wait a minute, you may ask, the loopback interfaces are not directly connected - they are NOT on the same data link, so how is each router going to know how to reach the NON-directly connected neighbors. After all, there's no IGP running between those routers. The answer is static routes. Using static routes that point to the directly connected WAN interface as a next hop, inform each router that the way to reach the BGP neighbor is through the WAN links.

So, the load balancing really becomes a function of basic routing and not so much a product of BGP. BGP would actually have gotten in the way of using those 2 WAN links had the loopback/eBGP multihop solution not been available.

Study this example:

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00800c95bb.shtml#ebgpmulithoploadbal

Understand?

Victor

wgranada1 · ‎05-12-2008

Hi Victor thanks for the info I will look at the example and the link you provided and let you know. Thank you for taking the time out to try to explain, I'll let you know

wgranada1 · ‎05-12-2008

Thank you everyone for your input and I just want to apologize for wasting your time. I just found out that the distant end has BGP disabled and that is why BGP never connected.

I appreciate you all taking the time to help me out!!!! Thank you and sorry about that

lamav · ‎05-12-2008

Its OK, dont be sorry, Just be grateful by giving us 5 points each for our posts. :-)

VL