Hello! I am working on a Project to implement a second transitt provider to get conectivity to the “world” I already have an MPLS cloud, using confederation and an internal ospf Routing table to distribute loopback for bgp peering to other routers to distribute my internal vrf - internet. This is ok, and working. At “end” i have a core router, having full vrf and managment (including loopback) table, ande use a default route from this to a CCR having a global Routing table Connected to my transit provider 1 (BGP). All Works like expected.
Now i try to implement a second transitt provider, conected to “other end” of the clod. “So far” so good. I setup a router and get a default route trough this provider ( not getting a full bgp table). Now I have some questions regarding best practic to implement this. All doc i have found in wiki etc, speak about using the same router to multihome bgp to the world. I cannot do this. i need this to be at 2 routers. Does annyone have a best practice here? I have it up running, but i feel like i loose some control of the Routing. I try to use local pref, as-path prepend etc, but still, i feel like i loose Control of the routes. Does annyone have comments to this setup? Should i have a direct BGP trough my 2 routers having default GW? should i let the vrf inside MPLS cloud just “go Cracy” With my routes, or does annyone have any other ideas? I want it to be as dynamic as possible, but when i use as less filters as i want and have all dynamic, some routes goes to one direction, some to the second, but mainly TX and RX goes and comes from one router eatch. Eg mainly traffic goestroug provider1 and come back trough provider 2, but its not consistant, and some goes at same" I can chooe by dissable default route at both ends, and have one or the other Active. But when i activate both, i just loose the Control, and canot tell exacty where the routes will goes, and why. eg why does the routes come back from provider1 when default gw is to provider 2, why does traffic goes out by provider 2 and come back trough provider 1 etc. And why dont all customer goes the same way? Some goes to provider 1 and some to provider 2
Hi
I have the same scenario, hoping for an answer!
Hi guys, there will be also the issue with incoming traffic.
It’s easy to change policy for outgoing BGP traffic, weight, local pref etc, but with flow of incoming traffic from source to your network is not such easy. You do not control this path, even AS-PATH prepending should not help you in some scenarios.
The way how to manipulate incoming BGP traffic is to play with splitting network to several /24s + setting of proper as-path prepending per this sub-networks or you can use BGP communities of ISP if he provides it.
I’m also running a similar setup, though I have mixed routers Mikrotik and Cisco 7206, and I don’t run MPLS.
But it does not matter as both BGP and OSPF are standards based.
So back to your setup…
Regarding your questions, what is recommended:
- Use IBGP between all of your edge routers.
- In that setup do not use a BGP session where one of the upstream routers is advertising only a default route. This is how you will not be able to take advantage of Local Preference. The optimal setup is to get the full (or at least partial) routing table from all ISPs. This will give you more precise control on how your traffic is exiting your AS, as already pointed out in this thread.
- If you advertise your own prefix to the rest of the world, as already mentioned here - you don’t have strict control on how the traffic from the outside of your AS. Yes, there are some BGP attributes you could use - but there are more or less not effective in a real world scenario. You could try AS path prepending, it might work to some extend.
Cheers,
Boyan