18.3.1 Example of system configuration of the mutual node switching system
The following figure shows an example configuration for a mutual node switching system that is integrated with cluster software. From outside the server such as from the client or database, Application Server appears to restart with the same logical address during the node switching.
|
|
![[Figure]](GRAPHICS/ZU180100.GIF)
The mutual node switching system shown in the figure is as follows:
-
Each Application Server of the executing node 1 (standby node 2) and standby node 1 (executing node 2) is managed in a different management domain.
-
In each management domain, two virtual hosts of the active node and spare node are defined. The virtual hosts of the active node and spare node are used as the executing node Application Server and the standby node Application Server respectively.
-
When failure occurs, the virtual hosts in the mutual management domains are switched. For example, as shown in the figure, if a failure occurs in the active node virtual host of server 1, the node is switched to the spare node virtual host of server 1.
-
The IP address to be used in the operations of each server is the IP address dynamically allocated by the cluster software (alias IP address). For the mutual node switching system of the figure, 'xxx.xxx.1.111' and 'xxx.xxx.2.222' become the alias IP addresses. Note that the cluster software switches LAN for each IP address, and therefore, a unique value is allocated to the cluster IP address of server 1 and server 2. During node switching, the alias IP address is deleted in the active node virtual host, and the alias IP address is added in the spare node virtual host so the processing is continued.
-
To send requests from the Management Server to the Administration Agent, use the IP address that does not move to another node due to node switching (stationary IP address) is used. In the mutual node switching system of the figure, 'xxx.xxx.0.11' of the active node 1 and 'xxx.xxx.0.22' of the spare node 1 become the stationary IP addresses.
- Reference note
-
A virtual host is a configuration in which multiple different IP addresses can be allocated to one machine, and the machine can be used as multiple physical hosts. The virtual hosts in the same management domain can control operations such as starting and stopping Application Server using one Administration Agent, but the IP addresses used for operations are different, so the virtual hosts are handled as apparently different physical hosts.
You can execute the following types of operations in the mutual node switching systems:
- Using the shared disk device
-
The use of the shared disk device is different in the local transaction and the global transaction.
-
In a local transaction
The shared disk device is not required. In the local transaction, no session information is to be inherited between the executing node and standby node, so the shared disk device is not used.
-
In the global transaction
The shared disk device is required. The shared disk device is used for inheriting the transaction information such as the OTS status during the node switching.
-
- For JP1 integration
-
In a cluster software configuration, you can integrate the system with JP1.
For integrating with JP1, Application Server also requires JP1/Base. You must manage JP1 in the cluster software separately from Application Server.
- Integration with cluster software in a database server
-
You can also use a cluster software configuration in the database server. In this case, if only the virtual address (logical address) is recognized in Application Server, you need not be aware that the database server is using the cluster software.
- Application of the load balancer
-
Though not described in this system configuration example, you can also apply the load balancer by providing several Web servers in the same configuration. As a result, you can improve the reliability and operating rate of the Web server.
For details on the system configuration of the mutual node switching system, see 3.11.4 Configuration in which executing node and standby node of Application Server are mutually standby in the uCosminexus Application Server System Design Guide.