Hitachi

JP1 Version 12 for UNIX Systems JP1/Performance Management - Agent Option for Platform Description, User's Guide and Reference


7.4.2 Settings for collecting user-specific performance data

To collect user-specific performance data:

  1. Determine the information to be stored in fields.

  2. Create user commands.

  3. Set the scheduler to collect user-specific performance data periodically.

  4. Specify the settings for collecting information from the user data file.

The following subsections describe the steps in this procedure.

Organization of this subsection

(1) Determining the information to be stored in fields

The fields of a user-defined record store two types of information, key information and data information. You will need to consider what to store as key information and what to store as data information.

(a) Key information

A user-defined record for storing user-specific performance data is a multi-instance record in which one or more rows can be stored by one collection run. To identify each record instance in one user-defined record, key information must be set. If you specify multiple user-created data files in the jpcuser command, you must set key information that uniquely identifies each record instance across all of the specified files. The following table shows the types of key information.

Table 7‒20: Types of key information

Type

Field name

Explanation

Transaction type

Trans Type

Identifies the instance type.

Transaction key

Trans Data Key (numeric type)

Identifies each of the instances that have the same transaction type.

Trans String Key (string type)

The transaction type is used to identify the type of the performance data. For example, assume that information about a database is stored in one record and information about a Web server is stored in another record. In this case, you can use DATABASE and WEB as transaction types to indicate which type of information (information about a database or information about a Web server) is stored.

When there are multiple instances that have the same transaction type, the transaction key is used to identify each instance. If neither the Trans Data Key field nor the Trans String Key field is set, or the same value is set for multiple transaction keys, the record instances cannot be identified uniquely. As a result, the first instance of each record is used.

(b) Data information

As data information, user-defined records can store three types of numeric data (double, long, and ulong types), three lengths of string data, and time data. The number of data items that can be stored differs depending on the user-defined record. For numeric data of the PI record type, either average or cumulative can be selected as the consolidation rule.

Select the user-defined record to be used based on the performance data to be collected. Note that a user-defined record that can store a larger amount of information consumes a larger amount of memory and other resources. We recommend that you select the user-defined record whose size is the minimum necessary.

The following table shows the number of fields for each type of user-defined record.

Table 7‒21: Number of fields for each type of user-defined record

Record type

User-defined record type

Number of fields

Numeric data

String data

Time data

PD record type

User Data Detail (PD_UPD)

2 × 3 = 6

1 + 2 + 4 = 7

1

User Data Detail - Extended (PD_UPDB)

5 × 3 = 15

5 + 5 + 5 = 15

1

PI record type

User Data Interval (PI_UPI)

4 × 3 = 12

1 + 2 + 4 = 7

1

User Data Interval - Extended (PI_UPIB)

10 × 3 = 30

5 + 5 + 5 = 15

1

User Data Interval - Expanded n#1 (PI_XUIn#1)

60 × 1 = 60#2

1#3 + 2#4 = 3

1

#1

n denotes a number in the range from 1 to 5.

#2

These are all double type.

#3

This is the number of 128-byte character strings.

#4

This is the number of 64-byte character strings.

The following table shows the criteria for selecting the recommended user-defined record.

Table 7‒22: Criteria for selecting the recommended user-defined record

Will cumulative data be stored as the performance data?

Will many types of performance data be stored?

Recommended user-defined record

Yes

No

PI_UPI

Yes

Yes

PI_UPIB, PI_XUI1 to PI_XUI5

No

No

PD_UPD

No

Yes

PD_UPDB

(2) Creating user commands

User commands are scripts that are used to collect performance data to generate user-created data. You must code the scripts so that performance data is output in the format used for user-created data files.

For details about the format of user-created data files, see 7.4.6 Format of user-created data files.

Store the user-created data at a path that can be specified by the -file option when the jpcuser command is executed. For details, see the description of the arguments in 7.4.5 Format of the jpcuser command.

To verify the user-created data output by the user commands, execute the jpcuser command in the following format:

When the command is executed, the following debug log file is generated:

Use the debug log file to check for errors.

For details about the jpcuser command, see 7.4.5 Format of the jpcuser command.

(3) Setting a scheduler to collect user-defined performance data periodically

The following explains how to set up the functionality for periodically executing user commands, to periodically collect user-specific performance data.

To periodically collect user-specific performance data.

  1. Set up user record collection in PFM - Web Console.

    The execution interval for functionality for periodically executing user commands depends on the Collection Interval setting for each user record.

  2. Set the properties for functionality for periodically executing user commands in PFM - Web Console.

    In PFM - Web Console, set the following properties for each user record to run functionality for periodically executing user commands. The method for setting these properties is the same for PD_UPD, PD_UPDB, PI_UPI, PI_UPIB, and PI_XUI1 to PI_XUI5 records.

Note:

Note that, when you use PFM - Web Console to specify the settings for periodically collecting user-specific performance data, the service to be selected differs depending on whether a physical host or a logical host is used:

  • Physical host environment: UNIXphysical-host-name or UA1physical-host-name

  • Logical host environment: UNIXlogical-host-name or UA1logical-host-name

Figure 7‒5: Properties for functionality for periodically executing user commands

[Figure]

Table 7‒23: Setting properties for user records

Property

Value

Description

Default value

User Command Setting - User Command Execution Timing

After/Before

By using the functionality for periodically executing user commands, specify when to execute user commands.

  • After: Execute user commands after record collection.

  • Before: Execute user commands before record collection.

After

User Command Setting - User Command Timeout

Integer in the range from 1 to 86,400

If you select Before as the User Command Execution Timing property of the functionality for periodically executing user commands, specify the time (seconds)# after which the execution of user commands is discontinued.

#:

The specified time when the execution of user commands is discontinued must not affect the timing of other record collection processing.

5

User Command Setting - record-name - Execute

Yes/No

Specify whether to use functionality for periodically executing user commands.

  • Yes: Use the functionality

  • No: Do not use the functionality

No

User Command Setting - record-name - UserCommand

Absolute path

Specify the absolute path for user commands. The maximum length of the string that can be specified for an absolute path is 255 bytes. Alphanumeric characters and symbols except for the following characters can be specified:

| < >

Blank

Notes:

1. When the Execute property is set to Yes and the UserCommand property is blank, the KAVF10203-W message is output, and the user command is not executed.

2. If the specified user command does not exist, or the user command does not have execution permissions, the KAVF10013-W message is output.

3. When a logical host is used, you can use the UserCommand property to specify the path to the user commands located on the shared disk. If user commands are not located on the shared disk, you need to locate them at the same path for both the executing node and standby node.

4. If you select After as the User Command Execution Timing property, and if the executed user commands do not finish by the time the next record collection processing begins, no user command is executed.

Note

The UNIX cron command can be used to periodically collect user-specific performance data. UNIX provides the cron command, which can automatically execute a batch file or program at the specified time and interval. After creating a shell script that executes the user commands and then the jpcuser command, set the cron command so that the shell script is executed periodically.

(4) Specifying the settings for collecting information from the user data file

The user data file contains data that the jpcuser command has converted from user-created data into a record format that can be managed by PFM - Agent for Platform. The data in the user data file is stored in user-defined records every time PFM - Agent for Platform collects records. Make sure that PFM - Web Console is set so that PFM - Agent for Platform will collect user-defined records.

For details about how to collect records, see the chapter on Performance Management functionality in the JP1/Performance Management Planning and Configuration Guide.