Client Fail over to different DC


#1

We would like to have 3 DC in different regions in the world using XDR. Is it possible to configure the client to auto reconnect to a remote DC if the entire DC that he is connected is down. Is it supported by the client SDK? If not, what is the best practice to implement something like this?

In each region, we are planning to have 3 nodes of Aerospike, hundred of our back-end engines are connected to this local Aerospike cluster all and the engines in the region should be not interrupted in cased of the entire local cluster of Aerospike is down, the engines should continue working by reconnect to a remote DC (remote Aerospike cluster).


#2

You could configure the client with three hostnames, one for each DC. The client will try to connect to them in order, so perhaps you want to change the order of those hostnames depending on location, with your nearest cluster first, farthest cluster last. If it cannot connect to the first hostname it will continue to the next one, and once it establishes a connection that will become the seed.

If you want to know which cluster you’re actually connected to you’d use service info command. Info commands can be issued from the API or a tool such as asinfo.