Job Management Partner 1/Base User's Guide
This section provides some notes on using JP1/Base in a cluster system. These notes are divided into common guidelines for both Windows and UNIX, notes for Windows only, and notes for UNIX only.
- Organization of this section
- (1) Guidelines common to both Windows and UNIX
- (2) Notes concerning Windows only
- (3) Notes concerning UNIX only
(1) Guidelines common to both Windows and UNIX
- When setting up JP1/Base in a cluster system, make sure that you stop the JP1/Base services active on the physical host and existing logical hosts. If you do not stop the JP1/Base services before setting up JP1/Base, the services on the logical hosts will not operate properly. If this happens, recover by restarting the server.
- To issue events from a user application in a cluster system, use the jevsend command with the -s option specified. For the -s option, specify the event server name. This option allows issued events to be inherited from the primary node to the secondary node when a failover occurs.
- When node switching is enabled, JP1/Base does not support duplication of the event database and command execution log (ISAM) file. Use a mirror disk or RAID disk to ensure the reliability of the disk system.
- When using JP1/Base in a cluster system, specify sync for the options parameter in the event server settings file (conf). The OS normally stores data written from a program in buffer memory, and then writes it to the disk in order to improve performance. Therefore, if the system suddenly terminates because of a power failure or an error in the OS, any data not yet written to the disk will be lost. The event service suppresses this buffering to prevent data from disappearing. If you specify no-sync for the option parameter or specify neither sync nor no-sync, data might be lost.
- The more logical hosts you concurrently activate in a cluster system, the greater the system resources required.
- To run JP1/Base on both the logical and physical hosts in a cluster system, you must change the event service setting on the physical hosts to IP addressing. Edit the event server settings file (conf) on both the primary and secondary nodes, changing the address specified in the ports parameter to the local host name or to the IP address of the local host. The event service on the physical host is set by default to 0.0.0.0, but the event service with this address cannot be activated concurrently with the event service on the logical host. For details on the event server settings file, see Event server settings file in 14. Definition Files.
- When the authentication server is switched during a failover, the related programs behave as follows:
- JP1/IM
- A communication error occurs. Operations are restored after the failover.
You can avoid potential problems due to this behavior of JP1/IM and JP1/AJS2 by placing the authentication server on a host outside the cluster system.
- JP1/AJS
- A communication error occurs. The user must log in again after the failover.
- If you want to monitor files on a shared disk using the JP1/Base log file trapping function, ensure that the shared disk remains accessible while the files are being monitored. If you change the shared disk allocation during file monitoring, problems such as errors in the monitoring process and control failure in disk space allocation and deallocation could occur.
- To prevent data from being lost from the command execution log file (ISAM), specify ON for the -flush option of the jcocmddef command, enabling the command execution log to be written line-by-line to the disk. For details on the jcocmddef command, see the manual Job Management Partner 1/Integrated Management - Manager Command/Definition Reference.
(2) Notes concerning Windows only
- Always perform the operations from the primary node when you are specifying an authentication server or registering JP1 users on a logical host. Be sure to start the JP1/Base service on the logical host before you start JP1 user registration.
- When backing up the definitions on the primary node in a cluster environment, make sure that all the letters in the logical host name that you specify in the jbsgetcnf command match the case of the logical host name in the definitions.
If you make a mistake, delete the logical host and specify it again.
- Before you delete a logical host, first stop the services of JP1/Base and the programs that require JP1/Base (JP1/IM, JP1/AJS, or JP1/Power Monitor) running on that logical host. If you delete a logical host with a service still active, delete that service in either of the following ways:
- Recreate a logical host with the same name, and then delete it.
- Uninstall JP1/Base.
- If you create a logical host with the same host name as the local host, the JP1/Base Event service on the physical host will be deleted if you later delete that logical host. To restore the service, execute the following command:
jevregsvc -r- If you cannot start or stop the JP1/Base service, a JP1/Base process might not have completed. In this case, restart the system.
- Even if you are not using JP1/Base on a particular physical host, the JP1/Base LogTrap service is still required for processing by the logical host. If you are not using the JP1/Base Control Service, set auto startup for the JP1/Base LogTrap service.
- The startup control function is not available for the services running on logical hosts. The startup control function is only available for the services running on physical hosts. Use cluster software to control startup of services on logical hosts.
(3) Notes concerning UNIX only
- If you stop JP1/Base services that support logical hosts, a JP1/Base process might fail to stop. In this case, execute the jbs_killall.cluster command to forcibly stop the process. Note that this command should be used only when JP1/Base processes cannot be stopped with the normal procedure. For details on the jbs_killall.cluster command, see jbs_killall.cluster (UNIX only) in 13. Commands.
- In a cluster system that performs monitoring even during stop processing, modify the command that terminates JP1/Base (the event service and process management including user management), as follows:
cd /etc/opt/jp1base
cp -p jbs_stop.cluster.retry.model jbs_stop.cluster
All Rights Reserved. Copyright (C) 2009, Hitachi, Ltd.