Thrift API

When Thrift is used (use_cql is false), all hosts specified in dbhost are used in a round-robin manner. If a host fails, it is skipped and periodically retried. This achieves load balancing across a set of servers.

Alternatively, you can set dbhost to a primary Cassandra node or set of nodes and failover to a secondary list of nodes only the primary node(s) are all unavailable. To do this, specify one or more Cassandra nodes using the secondary_dbhost parameter. Example:

dbhost: 10.1.82.146

secondary_dbhost: 10.1.82.147, 10.1.82.148

This example tells Doradus to use 10.1.82.146 exclusively (perhaps because network-near) and failover to 10.1.82.147 and 10.1.82.148 only if the primary node becomes available.