1.4 Overview of operation in a cluster system
A cluster system consists of multiple server systems that are configured in such a manner that if a failure occurs on one of the servers, applications can continue on another server. In a cluster system, hosts are classified as follows:
If a failure occurs on the active server that is running applications, the applications can be inherited to the standby server. The function for inheriting applications in the event of a failure is called system switchover. A logical server that is the unit of failover during system switchover is called a logical host.
In a cluster system, the applications must be run in a logical host environment so that they can continue their processing even after system switchover. You can run applications that are to be run on a logical host on any server, regardless of the physical server being used.
A logical host consists of the components explained below. The applications that are run as daemons or services store data on the shared disk and use a logical IP address to communicate.
|
Component of logical host |
Description of component |
|---|---|
|
Daemon or service |
Daemons and services are the applications that are run in a cluster system, such as JP1/AJS and JP1/Advanced Shell. If a failure occurs on the active server's logical host, the daemons or services with the same names are started on the standby server's logical host. |
|
Shared disk |
This is a disk unit that is connected to both the active server and the standby server. If this disk stores the information that is to be inherited during system switchover (such as definition information and execution status), the standby server inherits the connection to the shared disk in the event of a failure on the active server's logical host. |
|
Logical IP address |
This is an IP address allocated while a logical host is running. If a failure occurs on the active server, the standby server inherits the same logical IP address allocation. This enables the client to use the same IP address as if the same server is always running. |
- Important note
-
In this manual, a logical server that is the unit of failover during system switchover is called a logical host. However, some cluster software products and applications use different terms, such as group or package. Check the appropriate terminology in your system by referencing your cluster software's documentation.
A logical server that is the unit of failover during system switchover is called a logical host, while the physical server is called a physical host.
The following figure shows accesses during normal operation and after system switchover.
|
|
![[Figure]](GRAPHICS/ZU010600.GIF)
The following explains the figure.
-
Accesses during normal operation
While the active server is running, the shared disk and logical IP address are allocated on the active server where the daemons or services are running.
-
Accesses after system switchover
If a failure occurs on the active server, the standby server inherits the shared disk and logical IP address and starts the same daemons or services that were running on the active server. Even though the physical server has changed as a result of system switchover, it looks to the client as if the server with the same IP address is running because the standby server inherits the shared disk and logical IP address.
To run JP1/Advanced Shell in a logical host environment, you must have a shared disk to store the data that needs to be inherited during system switchover and a logical IP address. If you will be using the user-reply functionality, you must also set up the cluster software so that it can control the start, stop, and operation monitoring of the user-reply functionality's management daemon or service.
A JP1/Advanced Shell that is running in a logical host environment can inherit the job execution environment from the active server to the standby server during system switchover by using data stored on the shared disk. This means that JP1/Advanced Shell must store the spool on the shared disk. Note that execution of a job that was executing at the time system switchover occurred does not continue.
For details about the JP1/Advanced Shell settings required for cluster operations, see 2.9 Running in a cluster configuration.