Nonstop Database, HiRDB Version 9 System Operation Guide
This section explains the procedures for changing the number of server processes. The following topics are covered:
The maximum number of server processes that can be started by HiRDB (maximum number of active processes) is determined by the HiRDB system definition operands shown in the following table.
Table 9-2 Operands for specifying the maximum number of active processes
Type of server | Definition type | Operand |
---|---|---|
Single server | System common definition | pd_max_users |
Front-end server | System common definition | pd_max_users |
Dictionary server | Dictionary server definition or server common definition | pd_max_dic_process#1 |
Back-end server | Back-end server definition or server common definition | pd_max_bes_process#2 |
#1: When the pd_max_dic_process operand is omitted, the value of the pd_max_users operand in the system common definition is assumed.
#2: When the pd_max_bes_process operand is omitted, the value of the pd_max_users operand in the system common definition is assumed.
The number of server processes generated during HiRDB startup (that is, the number of resident processes) is determined by the value of the pd_process_count operand.
If there will be many processing requests in the system, setting the number of resident processes to a large value will enable all processing to be started immediately. If the number of resident processes is set to a small value and more processing requests are issued than there are resident processes, HiRDB will process the requests by starting as many server processes as possible up to the specified maximum number of active processes. In this case, processing cannot be started until these processes have started. However, when the number of resident processes is set to a larger value, more memory is required in order to keep server processes always active. Therefore, the number of resident processes should be set to an appropriate value taking into account the concentration level of processing and the required memory size.
The following two methods are available for changing the number of server processes:
You can use the pdchprc command to change the maximum number of active processes and the maximum number of resident processes. However, changing the number of processes with the pdchprc command is temporary; such a change remains in effect only until HiRDB is stopped or until the pdchprc command is executed again.
You can also change the number of server processes by directly changing the values in the operands listed in Table 9-2 Operands for specifying the maximum number of active processes and in the pd_process_count operand. Note, however, that before you can change a value in an operand listed in Table 9-2 Operands for specifying the maximum number of active processes, you must either terminate HiRDB normally or perform a planned termination. If you change a value in an operand shown in Table 9-2 Operands for specifying the maximum number of active processes after HiRDB has been terminated forcibly or after HiRDB has terminated abnormally, an error will occur when you attempt to restart HiRDB and HiRDB will not restart. This does not apply to changing only the value in the pd_process_count operand.
Use of the system reconfiguration command (pdchgconf command) enables you to modify HiRDB system definitions while HiRDB is running. Note that HiRDB Advanced High Availability must be installed in order to use this command. For details about changing HiRDB system definitions while HiRDB is running, see 9.2 Modifying HiRDB system definitions while HiRDB is running (system reconfiguration command).
The pdchprc command can be used during HiRDB operation to change temporarily the maximum number of active processes and the number of resident processes. Examples are provided below.
After online applications have terminated, this example reduces temporarily the number of resident server processes during night-shift batch processing. Processing requests are issued concurrently during online application processing, but because there should be no such concurrent processing requests during night-shift batch processing, there is no need to keep many server processes active. It should be sufficient to activate only as many server processes as are required by night-shift batch processing.
The following figure shows the procedure for reducing the number of server processes for batch processing.
Figure 9-2 Reducing the number of server processes for batch processing
In a mutual system switchover environment, it is possible after system switchover occurs for multiple HiRDBs to be running on a single server machine, resulting in a temporary increase in memory usage. An increase in memory usage of this type can be minimized by using the pdchprc command to reduce the number of HiRDB server processes.
The following figure shows the procedure for achieving reduced system operation in a mutual system switching environment.
Figure 9-3 Reduced system operation in a mutual system switching environment
Problems might result if the pdchprc command is used to set the maximum number of server processes to 0. For example, when the maximum number of server processes is set to 0 for a front-end server that is located in the same unit as the system manager, the following utilities can no longer be executed:
All Rights Reserved. Copyright (C) 2011, 2015, Hitachi, Ltd.