We're in the process of designing a new iteration of our network where we improve resilliency by adding a second datacentre.
We'll be adding a second datacentre, with an identical configuration of servers as our primary location. To achieve network connectivity, we're looking into a couple of possible methods.
See earlier questions http://serverfault.com/questions/86736/best-way-to-improve-resilience and http://serverfault.com/questions/101582/dns-round-robin-failover-and-load-balancing
I'm pretty convinced that BGP is the right way to go about this, and this question is not about RRDNS.
1) If we have 2 locations, do we announce the same IP address block from both locations?
2) If we did this, but had a management ssh interface on x.x.x.50 from datacentre A, but it was on x.x.x.150 in datacentre B.
What is the best practice mechanism for achieving this? Because if I were nearest to A, then all my traffic would go to x.50, but if i attempted to connect to x.150, I'd not be able to connect, because this address wouldn't be valid at A, but only at B.
Is the best solution to announce 2 different netblocks, one at each location, facilitating the need for RRDNS, or to announce a single block, and run some form of VPN between the two sites for managment traffic?
