4.3.12 Operands related to system log files
- 38) pd_log_dual = Y | N
- Specifies whether or not dual system log files are to be used.
- Y: Use dual system log files.
- N: Do not use dual system log files.
- Advantages
- When dual system log files are used, HiRDB collects the same system log information in both files, which are called File A and File B. If a failure occurs in one of the files while the collected system log file is being loaded, the file can be loaded from the other file, resulting in higher system reliability.
- Relationship to other operands
- When use of dual system log files is specified, the name of the File B system log file must be specified with the pdlogadpf operand in each server definition.
- 39) pd_log_remain_space_check = warn | safe
- Specifies the processing to be performed by HiRDB when the available space in the system log file falls below the warning level. This is called the facility for monitoring the free area for the system log file. For details on this facility, see the HiRDB Version 8 System Operation Guide.
- warn:
- When the available space in the system log file falls below the warning level, the KFPS01162-W message is output.
- safe:
- When the available space in the system log file falls below the warning level, scheduling of new transactions is suppressed, all transactions inside the server are forcibly terminated, and the KFPS01160-E message is output.
- Specification guideline
- Hitachi recommends the specification of safe because it can reduce the probability of abnormal termination of units due to system log file space shortage. However, when safe is specified, all transactions inside the server are forcibly terminated when a shortage occurs in the available space in the system log file. Therefore, the design of the system log file requires more accuracy. For details on system log file design, see the HiRDB Version 8 Installation and Design Guide.
- 40) pd_log_singleoperation = Y | N
- This operand is applicable when dual system log files are used; it need not be specified when dual system log files are not used.
- Specifies whether or not the single-operation mode is to be used for the system log files. Even if an error occurs in a system log file, making no dual system log files available, HiRDB (or a unit for a HiRDB/Parallel Server) can continue processing using the remaining single normal system log file without being abnormally terminated. This is called single operation of the system log files.
- The mode in which continuation of processing is permitted only with both system log files available (normal operation mode) is called double operation of the system log files.
- Y: Use single operation of the system log files.
- N: Do not use single operation of the system log files. Both system log files must always be used.
- Condition
- Y must be specified in the pd_log_dual operand.
- 41) pd_log_rerun_reserved_file_open = Y | N
- Specifies whether or not a system log file is to be opened automatically.
- If no system log file that can be overwritten is available during a HiRDB (or unit) restart, HiRDB opens a reserved file (if one is available), makes it overwritable, and continues processing. This is called automatic opening of system log file.
- A reserved file is used in the following cases:
- Between the time of a restart and the time when the first synchronization point dump is collected
- When none of the opened file groups can be overwritten
- Y: Open a system log file automatically (open and use a reserved file).
- N: Do not open a system log file automatically (do not use a reserved file).
- Advantages
- When Y is specified, the unit can be restarted as long as a reserved file is available, even though no file that can be swapped in is available during the unit restart.
- However, if a file that is in unload wait status is available, the unit is stopped; once that file has been unloaded, the unit can be restarted.
- 42) pd_log_rerun_swap = Y | N
- Specifies whether or not the system log files are to be swapped during a HiRDB (or unit) restart.
- Y: Swap the system log files.
- N: Do not swap the system log files.
- Advantage
- When Y is specified, there can be a physical separation between the system log files used before and after a restart. Therefore, the system log file that was being used before the restart can be reused during server operation.
- 43) pd_log_swap_timeout = wait-time-for-completion-of-system-log-file-swapping
<unsigned integer>((1-32580))<<180>> (seconds)- Specifies in seconds the wait time during which swapping of system log files must be completed. If a swap of system log files is not completed within the specified amount of time, the unit terminates abnormally.
- Specification guidelines
- Normally, there is no need to specify this operand. If it takes a long time to swap system log files for a reason such as poor machine performance, you can specify this operand with a value that is greater than the default value. To detect errors or delays quickly during system log file swapping so that the unit can be terminated (such as because of a disk failure), reduce the value of this operand.
- 44) pd_log_unload_check = Y | N
- Specifies whether or not HiRDB is to check the unload status of system log files.
- Y:
- Check the unload status (normal operation).
- N:
- Do not check the unload status. A system log file is placed in swappable status, regardless of its unload status, when both of the following conditions are satisfied:
- It is in overwritable status
- It is in extraction completed status (HiRDB Datareplicator)
- In such a case, the system log file operation method is to release checking of unload status. For details about this operation method, see the HiRDB Version 8 System Operation Guide.
- Advantages
- Specifying N provides the following advantages:
- Operations are simplified because the system log file unload operation is eliminated.
- It is not necessary to provide files for storing unload files.
- Specification guideline
- N should be specified if the system log file will not be needed for database recovery (i.e., if recovery from a backup collection point will be sufficient).
- Notes
- The following points apply when N is specified:
- Database can be recovered only if backups have been made.
- If this option is specified when the system log file is required for database recovery, it will not be possible to recover the database.
- 45) pd_log_max_data_size = log-input/output-buffer-size
<unsigned integer>((32000-523000))<<400000>> (Bytes)- Specifies in bytes the size of the buffer to be used for system log input/output operations.
- Specification guideline
- HiRDB/Single Server
Change the specification value according to the tuning method.
- HiRDB/Parallel Server
Specify a value that satisfies conditional expression 1 below. If uap is specified in the pd_rpl_reflect_mode operand or a recovery-unnecessary front-end server is used, specify a value that satisfies both the conditional expressions below (1 and 2). To optimize the value, use the tuning method for the specification value.
Conditional expression 1: log input/output buffer length
a
a: 72 x total number of back-end servers + 1432
Conditional expression 2: log input/output buffer length
b
b: (Maximum number of back-end servers subject to update processing by a single transaction + 1) x 128 + 64
- Tuning the specified value
- A value (other than the default value) may need to be specified for this operand after the following types of statistics analysis utility statistical information related to system operation have been checked:
- Number of buffer sectors waiting for input/output (# OF BUFFER FOR WAIT I/O)
- Number of waits caused by lack of a current buffer (# OF WAIT THREAD)
If the number of waits caused by lack of a current buffer is not 0, increase the value of this operand.
- Relationship to other operands
- Use this operand and the pd_log_write_buff_count operand to determine the log output buffer size.
- If v6compatible or v7compatible is specified in the pd_sysdef_default_option operand, the default value for this operand is 32000.
- 46) pd_log_write_buff_count = log-output-buffer-sectors-count
<unsigned integer>((3-65000))<<10>>- Specifies the number of buffer sectors to be used for system log output.
- Tuning the specified value
- A value (other than the default value) may need to be specified for this operand after the following types of statistics analysis utility statistical information related to system operation have been checked:
- Number of times the buffer was full (# OF BUFFER FULL)
- Number of waits caused by lack of a current buffer (# OF WAIT THREAD)
- If either of these values is large, a larger value should be specified in order to improve throughput.
- Relationship to other operands
- Use this operand and the pd_log_max_data_size operand to determine the log output buffer size.
- If v6compatible or v7compatible is specified in the pd_sysdef_default_option operand, the default value for this operand is 3.
- 47) pd_log_rec_leng = system-log-file-record-length
<unsigned integer>((1024, 2048, 4096))<<4096>> (Bytes)- Specifies the record length for the system log files; the specifiable values are 1024, 2048, and 4096.
- The record length specified in the -l option of the pdloginit command should be specified for this operand.
- Notes
- If a value that is different from the record length specified by the -l option of the pdloginit command is specified for this operand, system log files cannot be opened.
- For details about how to modify the system log file record length, see the HiRDB Version 8 System Operation Guide.