Cluster index replication control

[responsys_cm]

If I have a cluster made up of indexers that comprise multiple physical racks and power circuits, what is the best practice for controlling replication so that if a circuit failure takes out an entire rack, I'm guaranteed to have a searchable copy of all buckets somewhere in the cluster?

Do I make each rack its own site in a multi-site cluster? If so, how do I control the affinity of the search heads so that all sites in the same data center are seen as "local"?

Thx.

C

[jkat54]

Each rack will be its own cluster, cluster master will control a multi site cluster. Search heads will connect to the cluster master and thereby be able to search all clusters. No search affinity is necessary if all clusters are in the same data center.