Notes on records
When you collect records, note the following.
- Organization of this page
Precautions for collecting performance data
The following provides precautions for collecting performance data.
About the registry
This note applies to only Hyper-V.
Operation of PFM - RM for Virtual Machine is only supported in an OS environment using the standard settings.
Performance data might not be collected correctly if special OS settings are specified in some way, such as using the registry editor to edit the registry directly. This is also true if the special settings have been published in the Microsoft Knowledge Base.
About collection of log data
PFM - RM for Virtual Machine stores log data for all monitoring targets or log data consolidated on a group agent basis in data files. A data file is created for each record. The maximum size of each data file is 2 GB. Therefore, if many hosts are monitored by a PFM - RM for Virtual Machine instance or if many instances are generated for a record, the Store database might be unable to store all of the log data.
In particular, log data consolidated by a group agent is likely to put pressure on the data file size. Therefore, if you use a group agent to perform monitoring, you must reduce the number of hosts monitored by the instance. Alternatively, you must use the LOGIF property to filter data that can be stored in the Store database.
For details about how to estimate the disk space requirements for the Store database, see A. Estimating System Requirements.
Notes on changing the system resources of a monitored host
The following provides notes on changing the system resources of a monitored host.
About performance data before and after system resources are changed
If the system resources of a monitored host are changed, continuity between data before the change and data after the change will be lost. Therefore, performance data before the system resources are changed and performance data after the system resources are changed must be handled separately.
Record generation result when data cannot be collected
The following explains the record generation result when data stored in a field cannot be collected.
Record is not generated
No record is generated in the following case:
-
When PFM - RM for Virtual Machine cannot collect the performance data stored in a field defined as an ODBC key field
Note on monitoring VMware environment
-
If a VMware instance is monitored that uses NFS or some other network file system as the storage system, the instance cannot be obtained by using the PI_VPDI or PI_HPDI record.
-
If the monitoring target is VMware ESXi, Lockdown Mode is disabled. If Lockdown Mode is enabled, the collection of performance data fails.
-
The virtual machine that has symbols such as / (slash), \ (backslash) and % (percent) in its name can not be monitored.
Note about virtual machine name for Hyper-V
-
If there are multiple virtual machines with the same name in the same Hyper-V system, PFM - RM for Virtual Machine might not be able to collect the correct information. Make sure that each virtual machine has a unique name.
Notes on collecting KVM and Hyper-V records
For records of the PI record type, there are fields that require performance data to be collected at least twice.
Causes of the value of a field becoming 0
The value of a field becomes 0 in the following cases:
-
The field requires performance data to be collected at least twice, but performance data has been collected only once.
-
The field requires performance data to be collected at least twice, but performance data collection order is not chronologically correct.
Notes on collecting KVM records
-
When the monitoring target is KVM, let PFM - RM for Virtual Machine refer the mounted remote file system (such as the df command be able to executed correctly). During the mounted remote file system does not reply, if you specify Disk_Category of an instance "Y", Remote Agent service cannot collect the performance data correctly. In this case, perform the following procedure:
-
Change Disk_Category of an instance "N".
-
Stop the df process on the remote host by using the following command:
# kill -TERM(or KILL) "process ID of df"
-
Recover the remote file system to be mounted correctly such as restarting the NFS daemon.
-
Change Disk_Category of an instance "Y".
-
-
If the monitored host is KVM, the value of the Used field in the Host Logical Disk Status (PI_HLDI) record does not include the free space of reserved blocks managed by Linux file system. Therefore, the value of the Size field is not equal to the total of the Used and Free fields.
Notes on collecting Docker environment records
For records of the PI record type, there are fields that require performance data to be collected at least twice.
Causes of the value of a field becoming 0
The value of a field becomes 0 in the following cases:
-
The field requires performance data to be collected at least twice, but performance data has been collected only once.
-
The field requires performance data to be collected at least twice, but start time of collected Docker containers has been changed.
Notes on collecting Podman environment records
For records of the PI record type, there are fields that require performance data to be collected at least twice.
Causes of the value of a field becoming 0
The value of a field becomes 0 in the following cases:
-
The field requires performance data to be collected at least twice, but performance data has been collected only once.
-
The field requires performance data to be collected at least twice, but the container is restarted before the latest performance data is collected after the previous performance data was collected.
Note on monitoring the environment with logical partitioning feature
When the HBA and an NIC in dedicated mode are assigned to a logical partition, the following information cannot be collected:
-
PI_VPDI (VM Physical Disk Status)
-
PI_VNI (VM Network Status)
Other notes
-
The values of PI_V*** records are obtained from a virtual machine and are different from the performance data of the running guest OS actually.
-
"PerfNet ID: 2006" is recorded in the event log of the monitored host when monitoring the Hyper-V environment where there are 32 or more CPUs on the monitored host, but there is no problem collecting performance data. Also, ignore the event of "PerfNet ID: 2006".
-
If you specify the character strings other than ASCII to the image file name (and path name) of the guest OS, the value of the Disk ID field in the PI_VPDI record might not be displayed correctly.