Multi-broker replication domains

A multi-broker replication domain is a collection of replication entries, or replicator instances, used by clusters or individual servers within a cell. Multi-broker replication domains were created with a previous release of the product.

After we upgrade the deployment manager to the latest version of the product, we can create data replication domains only. Any multi-broker domains that we created with a previous release of the product are still functional, however, we cannot create new multi-broker domains or replicator instances with the administrative console.

A replication entry, or replicator, is a run-time component that handles the transfer of internal product data. All replicators within a replication domain connect with each other, forming a network of replicators.

Components such as session manager and dynamic cache can connect to any replicator within a domain to receive data from their peer components on other application servers connected other replicators in the same domain. If the replicator that a component is connected to fails, the component automatically attempts to reconnect to another replicator in the domain and recover any data that was missed while the component was not connected to a replicator.

The default is to define a replication domain for a cluster when creating the cluster. However, replication domains can span across clusters.

Global default settings apply to a given replication domain across a cell. ost default settings tune and control the behavior of replicator entries in managed servers across the cell. Such default settings control the use of encryption or the serialization and transferring of objects. Some default settings tune and control how specific product functions, such as session manager and dynamic caching, leverage replication, such as session use of partitions.

For situations that require settings values other than the default, change the values for a given replication domain. Settings include various resource allocation, replication strategies, such as grouping or partitioning, and methods, as well as some security related items.

If we are using replication for HTTP session failover, we might also need to filter where the session replicates. For example, only replicate to two places out of many. The global default settings define the partition size or number of groups and the session manager settings define the groups to which a particular instance belongs.

Filtering is less important if we are using replication to distribute information on data that is no longer valid and actual cached data maintained by dynamic caching. Replication does not occur for failover as much as for data synchronization across a cluster or cell when we likely want to avoid expensive costs for generating data potentially needed across those various servers.

that we can filter or segment using multiple replication domains.

Data replication domains