cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2455
Views
0
Helpful
17
Replies

BGP HELP

Mavrick25
Level 1
Level 1

Hello Everyone..

I work for an ISP in EUROPE. I'm having a problem with the BGP Peering towards another ISP as well as the iBGP peering within our AS.

Its a fantastic BGP question as it incorporates many different questions regarding the functionality of BGP and how it operates.

The problem is that I can't describe it properly through a form...

I was wondering, if there is someone.. a network professional that would like to assist me... please leave me your email address and I will forward you a visio file with the topology and current operational status so we can discuss how resolve a major issue I'm having and that I can't resole on my own.

Thanks

Mav

1 Accepted Solution

Accepted Solutions

Mav,

to answer you question about what would happened if one link fails, i'd need more details... just to avoid wrong (misleading) help from my (our) part.

It all depends on your inter AS routing, physical and logical architecture.

For the documentation, here you go:

LOAD BALANCING USING BGP:

http://www.cisco.com/en/US/tech/tk365/technologies_configuration_example09186a00800945bf.shtml

BGP PREFIX SUMMARY (refer to Configure Aggregate Addresses ):

http://www.cisco.com/en/US/docs/ios/12_0/np1/configuration/guide/1cbgp.html?bcsi_scan_64A94748AE3CD026=0&bcsi_scan_filename=1cbgp.html#wp5014

HTH,

View solution in original post

17 Replies 17

nordick26
Level 1
Level 1

Hey Mav,

i can at least try to help you.

As e-mail, use my nick name + gmail(dot)com

Have a nice rest of the day

Im going to create a Visio file and send it the people that would like to help.

Should receive it shortly..

Mav.. Thanks

dclark
Level 1
Level 1

I can try helping as well.

dclark@ccbootcamp.com

Hello Everyone,

I would like to begin by apologizing and also begin by saying thank you for your assistance. Below you will find the problem I'm facing and difficulty I'm having when creating a fully functional and operation peering relationship between eBGP neighbors and iBGP neighbors.

In this scenario there are 2 different BGP relationships, an eBGP relationship and an iBGP relationship. I will begin by describing the eBGP relationship I have.

Below you will find the topology of our eBGP relationship. We have what we will call our traditional Router A and Router B. There routers are GSR routers, they are the Gateway out of our Autonomous System, they have the following characteristics.

1. They peer on the physical interfaces

2. In the BGP neighborship configuration there is a prefix-list that blocks all routes out of our AS that are not part of our ip address scheme, meaning we have summarized 6 x /15 routes, if there is a route that does not belong to the /15 route it is not passed along.

3. Router A and Router B receive the Full Routing Table from the our internet partners.

4. Note: The cloud represents the internet.

Now, an introduction into our Autonomous System, the GSR routers are connected to 2 Cisco 7000 series routers in a full-mesh as illustrated below.

1. Router A and Router B receive the FRT from our Internet Peer.

2. They DO NOT pass the FRT to router C or D, but rather only the default route via BGP.

3. ISIS is the routing protocol used to reach the loopbacks of this iBGP peering relationship.

4. Router C and D accumulates all the routes within our AS and advertise these routes via iBGP to Router A and Router B. So, Router A and Router B have the Full Routing Table as well as the routes of our internal AS.

This is were the confusion happens.

Above I wrote that there is a prefix-list on Router A and Router B that prevents all the routes that don't belong to the 6 x /15 routes out to the internet remember?

Anyways, on Router A and on Router B there are 6 static routes to the same 6 x /15 subnet pointing into our AS via static route to the physical interface on Router C the same thing from Router B to Router D's physical interface?? Why??

And when taken away does not announce our network anymore to our internet peer.

When this happens (Removing the static routes that point into our AS) Router A doesn't receive any download because it's not advertising the network anymore to our internet peer but the upload works perfectly fine. Our internet peer doesn't receive networks from Router A and passes all the traffic to Router B risking saturation.

After analyzing the situation further, it seems that the following happens at an iBGP level:

My problem is this:

Why aren't the routes being considered?

Why are the static routes pointing into our AS???,

Why are the static routes pointing into our AS being advertised to our internet peer and not the routes learned via iBGP from Router A and Router B??

They should be automatically learned and known!!

Why when I take away those static routes, our GSR doesn't advertise our Network anymore??

What should happen is the following.

The GSR should receive and consider all routes arriving from Router C and Router D and advertise these routes to our Internet peer via a summarized prefix.

The links should properly load-balance from a download point of view and be able to self-sustain itself at a routing stand point..

I hope I was clear.. please let me know if you need anymore information.. Please and thank you

mailaglady2
Level 1
Level 1

Giuseppe Larosa
Hall of Fame
Hall of Fame

Hello Mav,

you have the option to attach the visio file or jpeg version also here in the forums.

Hope to help

Giuseppe

ok

Giuseppe,

Let me know if you can see the file!

Hello Mav,

so, if i understand it well, the main problem/question you have is why those static routes.

The answer is, that this is the way of working of BGP.

If you want BGP to advertise any route (prefix), before advertising, BGP checks the routing table for the same prefix.

Example: you want to advertise (to external peer) 10.1.1.0/24, which is a summary of 4 subnet's you are using in your AS (10.1.1.0/26; 64/26; 128/26; 192/26). You're receiving this 4 prefixis via iBGP, but in other BGP can advertise this out of you AS, you need to have 10.1.1.0/24 in your routing table. This is why you need to use the static route (pointing back to router you get the more specific route from, or pointing to Null0).

So, hope this answers your question, and if you need some help with load balancing, please add some more details so that we can help you with the config.

Have a nice day,

If this is true then the current config is correct..

The only problem is that the static route is pointing to a physical interface..

What happens is that when that link goes down it doesn't advertise the route anymore because the link is inaccessable..

If I was to point the static routes to null0 would it properly balance over the 2 links??

Mav..

What does everyone else think??

Nordick,

You wrote:

"You're receiving this 4 prefixis via iBGP, but in other BGP can advertise this out of you AS, you need to have 10.1.1.0/24 in your routing table. This is why you need to use the static route (pointing back to router you get the more specific route from, or pointing to Null0)."

In the router bgp config, I have configured the 6 prefixes with the network command.. shouldn't that help announce the networks..

I've studied BGP but I don't remember ever hearing about "when announcing to an external peer, you need static routes pointing back to where it receives the specific routes"..

I could be wrong..

Mav

Mav,

to be more precise, not only to BGP peers, but to be able to advertise the prefix to ANY bgp peer, you need to have the same prefix already in routing table.

As you're getting MORE specific prefixes from your iBGP, you have two possibilities.

1. You can advertise this more specific routes with no additional config.

2. You can advertise a summary of these prefixes - but to do so, the same prefix as you're about to summarize (and advertise) need to be present in router's routing table. Without this, router won't advertise the prefix even you have configured "network X.X.X.X mask Y.Y.Y.Y" in your BGP process.

Concerning the Load Balancing, it's not so simple question (mainly if you want to achieve OUT and IN bound traffic load balancing)

Outbound LB can be achieved using BGP Load Balancing in your AS.

Inbound LB using MED. If not supported by ISP for whatever reason, you can use AS path prepending, as AS is WELL KNOWN MANDATORY attribute.

HTH

Nordick26,

Thank you so much for you help.

Tonight I will change the static routes..

Not to point to the physical interfaces but to point to a Null0 interface..

That way if one of the links fail, we should receive the problem we had before..

What do you think?

P.S. Is there any documenation you can pass to me? That would be great..

And thanks for all your help, I will let you know how it goes!

Mav,

to answer you question about what would happened if one link fails, i'd need more details... just to avoid wrong (misleading) help from my (our) part.

It all depends on your inter AS routing, physical and logical architecture.

For the documentation, here you go:

LOAD BALANCING USING BGP:

http://www.cisco.com/en/US/tech/tk365/technologies_configuration_example09186a00800945bf.shtml

BGP PREFIX SUMMARY (refer to Configure Aggregate Addresses ):

http://www.cisco.com/en/US/docs/ios/12_0/np1/configuration/guide/1cbgp.html?bcsi_scan_64A94748AE3CD026=0&bcsi_scan_filename=1cbgp.html#wp5014

HTH,

Review Cisco Networking for a $25 gift card