Configuring Java CAPS JBI Components for GlassFish Clustering

Component Support for Clustering

All Java CAPS JBI components are supported in a clustered environment. Some components work inherently in a clustered environment and some require special configuration, notably in the areas of load balancing and failover. Stateful service engines (SEs), such as BPEL SE and IEP SE, use a persistence database to track cluster instances and distribute work. Other components, such as the Scheduler BC, Data Mashup SE, and LDAP BC, work inherently in a clustered environment without needing to know about the different instances.

The BPEL SE and IEP SE provide additional support for clustering at a component level for backwards compatibility. This means that multiple SE instances can be deployed on one GlassFish server in a non-clustered environment. This allows for persistence, high-availability, and failover without using a GlassFish server cluster for all components.

Load Balancing and Failover

Load balancing and failover functions depend on the individual components in the cluster. For example, load balancing is not applicable to Service Engines (SE), and is protocol specific for some of the Binding Components (BC). For example, the HTTP Load Balancer is required for SOAP transactions, while the File BC, JMS BC, and FTP BC all have built-in load balancing.

GlassFish provides additional support for high availability, load balancing, and failover with the HTTP Load Balancer plug-in. The HTTP Load Balancer distributes incoming HTTP and HTTPS transactions among the instances in a cluster and also fails over requests to another server instance if the original server instance becomes unavailable. The HTTP BC does not have its own load balancing capabilities, making the HTTP Load Balancer plug-in a useful tool when implementing the HTTP BC in a clustered environment. For more information, see Chapter 5, Configuring HTTP Load Balancing, in Sun GlassFish Enterprise Server 2.1 High Availability Administration Guide.

Load balancing and failover for the JMS BC is provided through a JMS broker cluster. For more information, see Chapter 10, Java Message Service Load Balancing and Failover, in Sun GlassFish Enterprise Server 2.1 High Availability Administration Guide.