Architecture of Server
Click to Enlarge
Configuration changes to the Linux kernel are required to increase the amount of shared memory available. HADB utilizes shared memory to communicate synchronously with the application server for performance and writes to disk asynchronously in a separate thread for recovery in the event of a failure.
Creating clusters utilizing DAS was a breeze. Instances can be configured and added to clusters even if they don't physically exist yet, and the DAS will automatically detect when the instance comes online. Configurations can be created and shared by clusters or copied to create similar configurations with the ability to tweak settings such as base ports, JVM settings, thread pools and logging details.
Once the cluster was configured it was necessary to configure the loadbalancer.xml file. This file specifies the members of the cluster and is utilized by Sun's Web server plug-in, which is supported by both Sun's ONE Web Server and Apache. In previous versions of Sun's application server it was necessary to manually edit this file, which meant continuing maintenance when adding or removing machines from the cluster. This functionality has been moved into the command line interface administrative tool, asadmin. The loadbalancer.xml file is now automatically generated in a few short asadmin commands and is subsequently updated automatically as instances are added or removed from the cluster.
Once the cluster and load balancing configuration was complete, I was able to easily start and stop instances from DAS. After starting both instances, I was able to deploy applications to the cluster in a few simple steps. Applications can be designated for deployment to any combination of clusters and individual server instances. Configuring JDBC connection pools and data sources was just as easy, and I was able to push the configuration to the cluster, so all machines in the cluster were automatically configured with the JDBC data sources in one easy step.