Notes on records
Note the following points when collecting records.
- Organization of this page
Notes on performance information when modifying system resources
Notes are provided below for modifying system resources using the DLPAR functionality of AIX, the vPars functionality of HP-UX, the DR functionality of Solaris, and the CPU hotplug functionality of Linux.
-
Regardless of whether a PFM - Agent for Platform service is running or stopped, if the system resources are modified, there is no continuity between the performance data before and after the modification. Therefore, the performance data before and after the modification must be treated as separate.
If necessary, back up the Store database before modifying the system resources, then clear the Store database and begin collecting new performance data. For details about how to back up the Store database, see the chapter explaining backup and restoration in the JP1/Performance Management User's Guide.
-
With respect to the DR functionality of Solaris or the CPU hotplug functionality of Linux, the performance data shown in the following table will not be collected correctly if system resources are modified while a PFM - Agent for Platform service is running. Therefore, before modifying the system resources, stop the PFM - Agent for Platform service. After the modification, restart the PFM - Agent for Platform service. For details about how to start and stop services, see the chapter explaining how to start and stop Performance Management in the JP1/Performance Management User's Guide.
The table below shows the types of system resources that are affected if they are modified while a PFM - Agent for Platform service is running, and the actions that you can take to correctly display reports after the modification.
Table 9‒9: System resource types and actions to be taken to correctly display reports after modification Record
OS and name of the function used to modify system resources
System resource type
Actions to be taken after modification
CPU - Per Processor Detail (PI_CPUP)#1
-
DLPAR of AIX
-
vPars of HP-UX
-
DR of Solaris
-
CPU hotplug of Linux
CPU
-
For a real-time report:
Display the report again.
-
For a historical report:
Specify the display period so that it does not include the date and time when the system resources were modified.
Device Detail (PI_DEVD)#1
DR of Solaris
Device
Device Summary (PI_DEVS)#1
Network Interface Detail (PI_NIND)#1
LAN board
Network Interface Summary (PI_NINS)#1
System Summary Overview (PI)#2
DLPAR of AIX
-
CPU
-
Memory
vPars of HP-UX
CPU
DR of Solaris
-
CPU
-
Memory
CPU hotplug of Linux
CPU
- #1
-
Fields other than those listed below are affected by a system resource modification:
• Interval (INTERVAL)
• Record Time (RECORD_TIME)
• Record Type (INPUT_RECORD_TYPE)
• Fields that are added only when data is recorded in the Store database
- #2
-
The table below shows the fields that are affected by a system resource modification.
Table 9‒10: Fields that are affected by a system resource modification System resource type
Target field
CPU
When modifying system resources using the DLPAR functionality of AIX, the vPars functionality of HP-UX, and the DR functionality of Solaris:
-
Active CPUs (NUMBER_OF_ACTIVE_CPUS)
-
CPU % (KERNELMODE_USERMODE_PERCENT)
-
Context Switches (CONTEXT_SWITCHES)
-
Context Switches/sec (CONTEXT_SWITCHES_PER_SECOND)
-
Idle % (IDLE_TIME_PERCENT)
-
Interrupts (INTERRUPTS)
-
Interrupts/sec (INTERRUPTS_PER_SECOND)
-
Kernel CPU % (KERNELMODE_PERCENT)
-
System Calls (SYSTEM_CALLS)
-
System Calls/sec (SYSTEM_CALLS_PER_SECOND)
-
Total Kernel-Mode Time (TOTAL_KERNELMODE_TIME)
-
Total Idle Time (TOTAL_IDLE_TIME)
-
Total User-Mode Time (TOTAL_USERMODE_TIME)
-
Total Wait Time (TOTAL_WAIT_TIME)
-
Traps (TRAPS)
-
Traps/sec (TRAPS_PER_SECOND)
-
User CPU % (USERMODE_PERCENT)
-
Wait % (WAIT_TIME_PERCENT)
When modifying system resources using the CPU hotplug functionality of Linux
-
Active CPUs (NUMBER_OF_ACTIVE_CPUS)
-
CPU % (KERNELMODE_USERMODE_PERCENT)
-
Idle % (IDLE_TIME_PERCENT)
-
Kernel CPU % (KERNELMODE_PERCENT)
-
Total Kernel-Mode Time (TOTAL_KERNELMODE_TIME)
-
Total Idle Time (TOTAL_IDLE_TIME)
-
Total User-Mode Time (TOTAL_USERMODE_TIME)
-
Total Wait Time (TOTAL_WAIT_TIME)
-
User CPU % (USERMODE_PERCENT)
-
Wait % (WAIT_TIME_PERCENT)
Memory
-
Alloc Mem % (ALLOCATED_MEMORY_PERCENT)
-
Alloc Mem Mbytes (ALLOCATED_MEMORY_MBYTES)
-
Effective Free Mem % (EFFECTIVE_FREE_MEM_PERCENT)
-
Effective Free Mem Mbytes (EFFECTIVE_FREE_MEM_MBYTES)
-
Effective Used Mem % (EFFECTIVE_USED_MEM_PERCENT)
-
Effective Used Mem Mbytes (EFFECTIVE_USED_MEM_MBYTES)
-
Free Mem % (FREE_MEMORY_PERCENT)
-
Free Mem Mbytes (FREE_MEMORY_MBYTES)
-
Total Physical Mem Mbytes (TOTAL_MEMORY_MBYTES)
-
-
Performance information about resources that are logically partitioned
When the Micro-Partitioning function is used in AIX, a record is created for each logically partitioned processor. Therefore, even when resources are part of the same physical resource, their performance is not affected by other logical resources. Note that for a single-instance record of the PI record type, the combined total value of all resources including logical resources is displayed.
The table below shows the records and fields that are affected by the Micro-Partitioning function.
|
Record name |
Field name |
|---|---|
|
CPU - Per Processor Detail (PI_CPUP) |
All |
|
Application Process Interval (PD_APSI) |
CPU % (CPU_PERCENT_USED) |
|
Application Process Overview (PD_APS) |
|
|
Process Detail (PD) |
|
|
Process Detail Interval (PD_PDI) |
|
|
Program Summary (PD_PGM) |
|
|
System Summary Overview (PI) |
Active CPUs (NUMBER_OF_ACTIVE_CPUS) |
|
CPU % (KERNELMODE_USERMODE_PERCENT) |
|
|
Idle % (IDLE_TIME_PERCENT) |
|
|
Kernel CPU % (KERNELMODE_PERCENT) |
|
|
Total Idle Time (TOTAL_IDLE_TIME) |
|
|
Total Kernel-Mode Time (TOTAL_KERNELMODE_TIME) |
|
|
Total User-Mode Time (TOTAL_USERMODE_TIME) |
|
|
User CPU % (USERMODE_PERCENT) |
|
|
Wait % (WAIT_TIME_PERCENT) |
|
|
Terminal Summary (PD_TERM) |
CPU % (CPU_PERCENT_USED) |
|
User Summary (PD_USER) |
|
|
Workgroup Summary (PI_WGRP) |
Record generation results when data cannot be collected
The following explains the record generation results when data to be stored in fields cannot be collected:
- Records cannot be generated
-
When PFM - Agent for Platform cannot collect performance data to be stored in the fields defined as ODBC key fields, no records are generated.
Information on terminated processes
Functions for collecting information on terminated processes cannot be used. Therefore, the performance data for the fields shown below cannot be collected.
|
Record name |
Field name |
|---|---|
|
Process Detail (PD) |
Accounting Flags (ACCOUNTING_FLAGS) |
|
End Time (END_TIME) |
|
|
Exit Status (EXIT_STATUS) |
|
|
Mem Charge (MEMORY_CHARGE) |
|
|
Process Detail Interval (PD_PDI) |
Accounting Flags (ACCOUNTING_FLAGS) |
|
End Time (END_TIME) |
|
|
Exit Status (EXIT_STATUS) |
Notes on reserved records
Under Record in the New Report > Field window of the PFM - Web Console report wizard, reserved records are displayed, and you can create a report of these records. You cannot, however, display a report of reserved records.
In addition, if PFM - Agent for Platform 07-00 or earlier is upgraded to 08-00 or later, records reserved in 08-00 or later appear during history collection setup. These records can be modified. However, if they are modified and then specified for data collection, no records will be collected. The following message is issued to the common message log during the first collection attempt only:
KAVF10600-W An invalid collection event occurred. (record=record-id, rc=maintenance-code)
- Note
-
When you perform a new installation of PFM - Agent for Platform 08-00 or later, the properties of the reserved records are not displayed.
Note on specifying a process name in an alarm condition
When you specify a process name in an alarm condition for the PD_APP or PI_WGRP record, only the string length that is within the range of data that can be collected for the PD record is used to determine the alarm condition. Accordingly, when the process name exceeds the data range as shown in the example below, and you specify *xyz in the alarm condition, the expected condition judgment cannot be performed, since PFM - Agent for Platform does not hold the xyz part. In this case, specify an alarm condition within the range of data that can be collected for the PD record by, for example, specifying abc*.
- Example:
abc... xyz (the xyz part is outside the range that can be collected for the PD record)
Notes on cases in which record instances cannot be uniquely identified
For the PD_UPD, PD_UPDB, PI_UPI, PI_UPIB, and PI_XUI1 to PI_XUI5 records, the first instance of each record is used.
Notes about records
-
When a value is not specified in the User Time 1 field of PD_UPD, PD_UPDB, PI_UPI, PI_UPIB, PI_XUI1, PI_XUI2, PI_XUI3, PI_XUI4, and PI_XUI5 record, it is displayed as "1970 01 01 09:00:00." in PFM - Web Console.
-
Although PFM - Web Console displays values after the decimal point for the User Long N# field and User Long Roll N# fields of PD_UPD, PD_UPDB, PI_UPI, and PI_UPIB records, only integer values can be specified.
- #
-
N is a user-chosen numerical value.
Other note
-
Language environments other than Japanese, English, and Chinese support only the ASCII character set. In such environments, non-ASCII characters in performance data are not displayed correctly.