We have a country wide ISP managed MPLS network which all our offices connect to, we want to have two data centres on either side of the country which will serve as the exit points for internet (also same ISP). ISP managed eBGP is already running over the MPLS network.
Note: the two new DCs do not have an interconnect between them, they will have to use the MPLS network to send traffic between them.
If we peer with the ISP eBGP network at each DC and we advertise a default route from each DC into BGP, then the default route in the routing table at each office should choose the default route of the closest DC (the shortest AS path).
I'm just going by the BGP route selection criteria, and as the default routes propagate through the MPLS network, each router should choose the default route with the shortest AS path.
We have no weight set, no local pref. set, and both DCs would be redistributing at static default route, so AS path should be the deciding factor. Then if both AS path lengths are the same, then the next likely criteria would be the oldest route, which should be fine for tie-breaker.
So at the MPLS PE for each office there will be two default routes in the BGP table and the one with the shortest AS path will be the one in the routing table, or the one with the oldest route.
I should hopefully be able to verify this when we start peering with the MPLS PE at the DC and check the AS path from different locations. I can also run some tests with a shared dummy network.
We can then migrate from our current DC by advertising both the two new DCs (for a total of 3 default routes), and then remove the old DC when the two new ones are ready.
Is my thinking correct?
No comments:
Post a Comment