To use PFM - Agent for Service Response in a cluster system, the following prerequisites must be satisfied.
(a) Cluster system
Make sure that the following conditions are satisfied:
- The cluster system is controlled by cluster software.
- The cluster software is set up in such a manner that it controls startup and termination of the PFM - Agent for Service Response that is operating on the logical host.
- Notes:
- If a message box is displayed for an application error in Dr. Watson's log, failover may fail. In such a case, you must suppress the error notification that is displayed in this message box. For details about the suppression procedure, see the documentation that comes with the OS. Note that suppression of error notification may have adverse effects on information acquisition in the event of application errors.
- In Windows Server 2003, a dialog box for reporting an error to Microsoft is displayed when an application error occurs. This dialog box may cause failover to fail; therefore, reporting of the error must be suppressed. For details about the suppression procedure, see the documentation that comes with the OS.
- If you upgrade from a version 08-00 of PFM - Agent for Service Response to a newer one while it is operating in a cluster system, you must copy the measurement condition registration file on the shared disk in advance. For details about, see 5.3.4 Setup.
(b) Shared disk
Make sure that the following conditions are satisfied:
- A shared disk is available to each logical host and information can be inherited from the executing node to the standby node.
- The shared disk is physically connected to each node by means of a Fibre Channel or SCSI.
Performance Management does not support a configuration in which a disk replicated by a network drive or via a network is used as a shared disk.
- In the event of a failover, the shared disk can be forced offline (for example, by the cluster software) to achieve failover even if some processing using the shared disk continues due to a problem.
- Each instance of Performance Management programs on the same logical host uses the same directory on the shared disk.
Note that you can change the location of the Store database to another folder on the shared disk.
(c) Logical host names and IP addresses
Make sure that the following conditions are satisfied:
- Each logical host has a logical host name and a corresponding logical IP address, and this information can be inherited from the executing node to the standby node.
- The logical host names and logical IP addresses are set in the hosts file and name server.
- If DNS operation is used, the host name without the domain name is used as the logical host name, not the FQDN name.
- All physical and logical host names are unique throughout the system.
- Notes:
- Do not specify a physical host name (that is, a host name displayed by the hostname command) as a logical host name. If it is specified, normal communication processing may not be achieved.
- A logical host name must consist of 1 to 32 bytes of alphanumeric characters. None of the following symbols or the space character can be used:
/ \ : ; * ? ' " < > | & = ,.
- A logical host name cannot begin with localhost, an IP address, or the hyphen (-).
(d) Settings when IPv6 used
Because Performance Management supports network configurations for IPv4 environments and IPv6 environments, you can use Performance Management in a network configuration that contains both IPv4 and IPv6 environments.
PFM - Agent for Service Response can communicate with PFM - Manager by using IPv6.
Note, however, that this applies only when Windows Server 2008 R2 runs on a host with PFM - Agent for Service Response installed and Windows Server 2008 R2 or Linux runs on a host with PFM - Manager installed.
For details about the scope of communication in the IPv4 and IPv6 environments, see L. About Communication in IPv4 Environments and IPv6 Environments.
To use IPv6 for communication, IPv6 must be enabled on both the PFM - Manager host and the PFM - Agent host by using the jpcconf ipv6 enable command. The following describes when you need to execute this command.
- Cases when the jpcconf ipv6 enable command needs to be executed:
- When switching from an IPv4 environment to an IPv6 environment on each host
- When switching the PFM - Manager environment from IPv4 to IPv6 in an environment in which both IPv4 and IPv6 are used
- Cases when the jpcconf ipv6 enable command does not need to be executed:
- When each host already operates in an IPv6 environment
- When PFM - Manager operates in an IPv6 environment in an environment in which both IPv4 and IPv6 are used
The following shows an example of executing the jpcconf ipv6 enable command:
jpcconf ipv6 enable
Execute the jpcconf ipv6 enable command on both the executing node and the standby node. For details about the conditions or occasions for executing the jpcconf ipv6 enable command, see the chapter that describes network configuration examples in an environment that includes IPv6 in the Job Management Partner 1/Performance Management Planning and Configuration Guide.
To set up PFM - Agent for Service Response for logical host use, you must provide the information listed in the table below, in addition to the environment information that is needed to set up a normal PFM - Agent for Service Response.
Table 5-2 Information required to set up PFM - Agent for Service Response for logical host use
Item | Example |
---|
Logical host name | jp1-halsr |
Logical IP address | 172.16.92.100 |
Shared disk | S:\jp1 |
When multiple Performance Management programs are installed on a single logical host, the programs share the same folder on the shared disk.
For details about the capacity required on the shared disk, see A. Estimating System Requirements.
If you employ a system configuration in which PFM - Agent for Service Response operates on a logical host, you should evaluate whether or not the entire logical host should be subject to failover in the event of a PFM - Agent for Service Response failure.
If a PFM - Agent for Service Response failure results in failover of the entire logical host, a job application that is running on the same logical host will also result in failover, which will affect the job.
Hitachi recommends that you use one of the following cluster software settings so that errors on PFM - Agent for Service Response do not affect system operations on the same logical host:
- Do not monitor operation of PFM - Agent for Service Response
- Do not allow detection of errors in PFM - Agent for Service Response errors to result in failover