17.6.2 Collecting data if a problem occurs in UNIX (except for PFM - Web Console)

(1) Executing the data collection command

Use the jpcras command to collect the data needed to determine the cause of an error. Note that the user who executes the procedures described below must be a root user.

To execute the data collection command:

1. Log on to the host where the service subject to this data collection is installed.

2. Specify in the jpcras command the data to be collected and the storage directory for the data, and then execute the command.

   The following shows an example of specifying the command when all information that can be obtained by the jpcras command is to be stored in the /tmp/jp1pc/mgr directory:

   jpcras /tmp/jp1pc/mgr all all

   The data collected by the data collection command is compressed by the tar command and the compress command or the gzip command, and then stored in the specified directory. Example of the file name:

   Data collected by the jpcras command: jpcrasYYMMDD.tar.Z

   YYMMDD represents the year, month, and date.

When the jpcras command is executed, the jpctool service list -id "*" -host "*" command is executed internally to collect a list of PFM services and check their operating status. If there is a firewall between the host executing the command and a host on a different PFM system, or if the system is large, it may take an extended period to complete the jpctool service list -id "*" -host "*" command. In such a case, by setting the value of the JPC_COLCTRLNOHOST environment variable to 1, you can suppress the jpctool service list -id "*" -host "*" command, to reduce the time required to complete the command.

For details on the jpcras command, see the chapter explaining the commands in the manual JP1/Performance Management Reference.

To Page Top

(2) Executing the data collection command (for logical host use)

The data of Performance Management for logical host use exists on the shared disk, and this data must be collected from both the executing node and standby node.

Use the jpcras command to collect the data needed to determine the cause of an error. The following describes the procedure for executing the data collection command. Note that the user who executes the procedure described below must be a root user.

To execute the data collection command for logical host use:

1. Mount the shared disk.

   The data of the logical host is stored on the shared disk. For the executing node, make sure that the shared disk is mounted and then collect the data.

2. For both the executing and standby nodes, specify in the jpcras command the data to be collected and the storage directory for the data, and then execute the command.

   The following shows an example of specifying the command when all information that can be obtained by the jpcras command is to be stored in the /tmp/jp1pc/mgr directory:

   jpcras /tmp/jp1pc/mgr all all

   The data collected by the data collection command is compressed by the tar command and the compress command or the gzip command, and then stored in the specified directory. Example of the file name:

   Data collected by the jpcras command: jpcrasYYMMDD.tar.Z

   YYMMDD represents the year, month, and date.

   Point:

   To determine the cause of an error in the logical host environment, you need to obtain the data about Performance Management on both physical and logical hosts. Execute the jpcras command without the lhost argument on a node connected to the shared disk. You can collect all the data about Performance Management on both physical and logical hosts for that node. For this reason, do not specify the lhost argument in the jpcras command when Performance Management exists in the logical host environment.

   If the jpcras command is executed on a node in which the shared disk is not mounted, files on the shared disk cannot be acquired, but the command ends normally without an error.

   Note:

   Execute the command to collect the data on both the executing node and standby node. To investigate the conditions before and after a failover, the data of both the executing node and standby node is required.

3. Collect the cluster software data.

   This data is required to investigate whether an error occurred in either the cluster software or Performance Management. Collect the data to enable an investigation of the control request, such as a start or stop request from the cluster software to Performance Management, and the results.

When the jpcras command is executed, the jpctool service list -id "*" -host "*" command is executed internally to collect a list of PFM services and check their operating status. If there is a firewall between the host executing the command and a host on a different PFM system, or if the system is large, it may take an extended period to complete the jpctool service list -id "*" -host "*" command. In such a case, by setting the value of the JPC_COLCTRLNOHOST environment variable to 1, you can suppress the jpctool service list -id "*" -host "*" command, to reduce the time required to complete the command.

For details on the jpcras command, see the chapter explaining the commands in the manual JP1/Performance Management Reference.

To Page Top

(3) Checking information about the operation

Check the information about the operation when an error occurs and record the information. You also need to check the following information:

• Details of the operation

• Time the error occurred

• Machine configuration (for example, the OS version, host name, and PFM - Manager and PFM - Agent or PFM - RM configuration)

• Whether the error is replicable

• When logged on from PFM - Web Console, the Performance Management user name at logon

To Page Top

(4) Collecting error information

You need to obtain the following error information:

• Messages output to the console, if the error occurred during command execution

To Page Top

(5) Collecting other information

You also need to collect the following information:

• The command arguments that were specified, if the error occurred during command execution

To Page Top