This subsection describes the system configuration for switching nodes when cluster software is used and an error occurs.
In Application Server systems, the respective OS uses the following cluster software:
- In Windows
Windows Server Failover Cluster
- In AIX, HP-UX, or Linux
HA monitor
- In Solaris
The system linked to the cluster software cannot be operated.
In systems configured with Application Server, the following configurations can be realized using cluster software. Note that these configurations must be operated using Management Server.
- Configuration where executing node and standby node of Application Server are in 1-to-1 ratio
This configuration consists of one Application Server machine in the standby node for one Application Server machine in the executing node. If an error occurs on the Application Server machine in the executing node and the machine stops or if Administration Agent ends, the cluster software starts the Application Server machine in the standby node, and the processes are switched. Whether a shared disk device is necessary depends on whether you use a transaction service. Note that when you use Windows Server Failover Cluster, you need a shared disk device even if a transaction service is not used.
- Configuration where executing node and standby node of management server are in a 1-to-1 ratio
This configuration consists of one management server machine in the standby node corresponding to one management server machine in the executing node. If an error occurs on the management server machine in the executing node, and the machine stops or if the Management Server process ends, the cluster software starts the management server machine in the standby node, and the processes are switched.
- Configuration in which the executing node and standby node of Application Server are at mutual standby
This is one of the configurations where the executing node and the standby node of Application Server are in 1-to-1 ratio. Deploy a J2EE server of the same type on each Application Server machine, and start a different J2EE server. Each Application Server instance will function as the standby node for each other while operating as the executing node. If an error occurs in any of the nodes, the nodes are switched. As a result, operations with very little wastage are possible on a few Application Server machines.
With this configuration, Management Server is started on each Application Server machine, and their respective different management domains are controlled.
- Configuration where one standby node is used for N executing nodes (configuration in which a recovery server is used)
If a global transaction is used in a configuration where J2EE servers are redundant, this system configuration is used to conclude transactions when an error occurs on a specific J2EE server.
Use a load balancer to make N J2EE servers redundant.
- Configuration where the executing node and standby node of the host unit management model have an n-to-1 ratio
This system is used for allocating many executing nodes (1-to-N machines) of Application Server (Host) instances and one machine in the standby node, and allocating the respective Management Server and Administration Agent instances. When there is a failure on an Application Server machine in the executing node, you can continue the business by performing node switching to the Application Server machine in the standby node.
- Organization of this section
- 3.11.1 Configuration in which executing node and standby node of Application Server are in 1-to-1 ratio (when transaction service is not used)
- 3.11.2 Configuration in which executing node and standby node of Application Server are in 1-to-1 ratio (when using transaction service)
- 3.11.3 Configuration in which executing node and standby node of Management Server are in 1-to-1 ratio
- 3.11.4 Configuration in which executing node and standby node of Application Server are mutually standby
- 3.11.5 Configuration using server exclusive for recovery (N-to-1 recovery system)
- 3.11.6 Configuration in which the executing node and standby node of the host unit management model are in an N-to-1 ratio