Neo4j 3.1 introduced support for a new type of clustering known as causal clustering. With causal clustering, a few core servers will be configured, along with many read replicas. The configuration for causal clustering is similar to HA clustering:
causal_clustering.initial_discovery_members=192.168.0.100:5000,192.168.0.101:5000,192.168.0.102:5000
causal_clustering.expected_core_cluster_size=3
dbms.mode=CORE
This will define the 192.168.0.100-102 instances as core servers. Note that dbms.mode is set to CORE. The ha.initial_hosts property is replaced with the causal_clustering.initial_discovery_members property. The expected number of core servers is defined with the causal_clustering.expected_core_cluster_size property.
Configuring a read replica requires the same three lines in the configuration file, except that the dbms.mode is specified differently:
dbms.mode=READ_REPLICA
Clustering (high-availability or causal) is not available in the Community Edition of Neo4j. If your application requires either HA or causal clustering, please reach out to Neo4j, Inc. for assistance with licensing and installing the Enterprise Edition.