I have attached a topology which we are looking to achieve for some routing design. Below are the questions:
1) We need that the P-2-P link between both the Data Center is used only by the traffic betweens the Data Centers.
2) Consider a scenario that the link between Provider-2 and the Data Center goes down. In this case Branch Office-02 should reach the Data Center via the Backup Data Center --> Provider-1 --> Data Center.
What are the option that I can enable this to work?
Assuming the provider networks are MPLS.
Are you already running EBGP between the DCs ?
If so you shouldn't have to do anything because DC1 (main DC) is advertising it's networks direct to provider 2 and also to DC2 (backup DC).
DC2 then receives DC1s subnets and also advertises them to provider 2. But DC2 will have added it's own AS into the path which means the PEs within the SP's cloud will prefer DC1, because the routes have a shorter AS path, unless it stops advertising it's own subnets.
Are the DCs using the same IP subnets, is that the issue ?
Sorry i didn't read the question properly.
I assumed you wanted traffic to go via the P2P if the MPLS link in DC1 failed but you don't.
So start again :-)
Are you running EBGP across the P2P link ?
Are all the routers within each DC IBGP peers ?
correct, other than the traffic between the DC no other traffic should go over the P-2-P link. :-)
yes, eBGP is running across the P-2-P link.
Routers withing the DC are iBGP peers.
Okay, then what i said before still stands in terms of route advertisements and path selection to the branch office.
Using your example you just need to make sure that within DC2 the next hop IP for DC1s subnets are the provider 1 router and not the P2P link.
As you indicated on your diagram local preference would be the way to do this within DC2.
Sincere apologies, this is not one of my better posts.
Are you advertising with EBGP provider 1's branch sites to provider 2 and the same the other way round ?
Okay, then I think using local preference is the answer.
As long as each DC has routes back to the other DC for the branches via both providers then routing should work.
I think I got a bit confused going through all the replies. :)
Sorry for that.
Local pref can be the answer but in this case Local pref will always take the path between the P-2-P link between the DC's., right?
The service provider routers at the Data Centers will still think that the P-2-P router is the best path and they would route traffic through that link.
No problem, I think it was my poor answers that created the confusion.
You can manipulate the local preference so for example on the provider 2 router in DC2 you would be receiving IBGP routes from the P2P router and the DC2 provider 1 router for DC1s subnets.
You need to manipulate local preference on the provider 2 router in DC2 so that it sees the best route to DC1 via the provider 1 router and not the P2P router.
That way if the provider 2 connection to DC1 fails then traffic from the branch comes in to DC2 is sent to the provider 1 router in DC2 then across the cloud to DC1 and return traffic comes back the same way and is sent back to the branch from the provider 2 router in DC2.
Does that makes sense ?
Consider, that I manipulated local preference on the provider 2 router in DC2 so that it sees the best route to DC1 via the provider 1 router and not the P2P router.
Now once the traffic reaches that provider 1 router routes will still go to the P-2-P router. This will be because the P-2-P router still has the best path and routes. :(
Also this will cause the DC1 to favor the providers to reach DC2 and not the P-2-P link.
Why would it go to the P2P link ?
Local preference is higher in the BGP best path selection than AS path so it should use the provider 1 cloud.
You would obviously need to manipulate the local preference on that router as well or you could use weight everywhere instead it's up to you.
Local Pref would still be taken into account by the entire AS i.e. all the routers in the iBGP, correct?
Assume I modify the Local Pref at the Data Center Provider routers. These routers would still see the iBGP neighbor advertising the better path to the Backup Data Center.
I think weight should do the trick here. Let me try that out in the GNS3.
I forgot about the fact that you still want traffic between the DCs themselves to use the P2P link and was only thinking of branch traffic to and from the DCs.
So yes I think weight may solve your problem rather than local preference as it is not passed to any other BGP peer.