5.2.17 Operands related to synchronization point dump files

53) pd_spd_dual = Y | N
Specifies whether to use dual synchronization point dump files.
If Real Time SAN Replication based on the log-only synchronous method is used, the synchronization point dump files for log application processing for the server are also used at the log application site.
Y: Uses dual synchronization point dump files.
N: Does not use dual synchronization point dump files.
Advantage
When dual synchronization point dump files are used, HiRDB collects the same synchronization point dump in both dump files A and B. Even when an error occurs in one of the files when the collected synchronization point dump is being loaded, the synchronization point dump can be loaded from the other file, thus improving system reliability.
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is N.
Relationship to other operands
To use dual synchronization point dump files, specify the name of synchronization point dump file B in the pdlogadpf operand.
54) pd_spd_assurance_msg = Y | N
Specifies whether the KFPS02183-I message is to be output when a synchronization point dump is completed.
Y: Output the message.
N: Do not output the message.
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is Y.
55) pd_spd_assurance_count = number-of-guaranteed-valid-generations
~<unsigned integer>((1-2))
Specifies the range of system log files to be saved during HiRDB operation as a protection against events such as a synchronization point dump file input error during system recovery. What is specified here is a number of synchronization point dump file generations; the specified value is referred to as the number of guaranteed-valid generations. The number of past generations of synchronization point dump files specified here are overwrite disabled.
If Real Time SAN Replication based on the log-only synchronous method is used, this operand is ignored, if specified, at the log application site. The system log files used at the log application site are overwritten by the transaction execution site even if they include system log information that has not been applied.
Advantage
When 2 is specified as the number of guaranteed-valid generations and an error occurs in the most recent synchronization point dump file generation, the system can be recovered using the preceding synchronization point dump file generation, resulting in higher reliability.
Specification guidelines
  • To improve reliability, specify 2 for the number of guaranteed valid generations. However, the number of overwrite disabled synchronization point dump files increases (to two).
  • If reliability has been improved with the use of dual synchronization point dump files, we recommend that you omit this operand or specify 1 for it.
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is 1.
Notes
  • The minimum number of synchronization point dump files needed is number-of-guaranteed-valid-generations + 1.
  • Specifying 2 increases the number of overwrite disabled synchronization point dump files. Moreover, the system log files that correspond to the overwrite disabled synchronization point dump files are also overwrite disabled. Consequently, specifying 2 for the number of guaranteed valid generations increases the number of overwrite disabled system log files. As a result, a shortage might occur in the number of system log files that can be swapped in. To prevent this, it might be necessary to re-evaluate the system log file capacity.
56) pd_spd_reduced_mode = reduced-mode-operation-option
~<unsigned integer>((0-2))
Specifies whether the reduced mode operation for synchronization point dump files is to be used.
Reduced mode operation is a facility for continuing processing if at least two synchronization point dump files can be used, even if an event such as a file error during HiRDB operation or restart reduces the number of synchronization point dump files to or below the number of required files (number of guaranteed valid generations# + 1).
#: Value specified for the pd_spd_assurance_count operand.
0: Do not use the reduced mode operation.
1: Use the reduced mode operation.
2: Use the reduced mode operation and issue a warning message whenever a synchronization point dump is being acquired during the reduced mode operation.
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is 0.
57) pd_spd_reserved_file_auto_open = Y | N
Specifies whether a synchronization point dump file is to be opened automatically. When Y is specified and an error in a synchronization point dump file reduces the number of synchronization point dump files to the number of files necessary for operation (number of guaranteed valid generations# + 1), HiRDB opens a reserved file (if one is available), makes it overwritable, and continues processing. This is called automatic opening of synchronization point dump file.
If Real Time SAN Replication based on the log-only synchronous method is used, the synchronization point dump files for log application processing for the server are also used at the log application site.
#: Value specified for the pd_spd_assurance_count operand.
Y:
Open a synchronization point dump file automatically. A reserved file is opened when the number of files falls below the number necessary for operation (number of guaranteed valid generations + 1).
N:
Do not open a synchronization point dump file. No reserved file is opened, even though the number of files falls below the number necessary for operation (number of guaranteed valid generations + 1).
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is N.
Relationship to other operands
This operand has a higher priority than the pd_spd_reduced_mode operand.
58) pd_spd_max_data_size = synchronization-point-dump-file-buffer-size
~<unsigned integer>((32000-4000000)) (bytes)
Specifies in bytes the size of the buffer (shared memory) to be used for synchronization point dump file input/output operations.
The value specified here affects the number of synchronization point dump file input/output operations.
If Real Time SAN Replication based on the log-only synchronous method is used, the synchronization point dump files for log application processing for the server are also used at the log application site.
Specification guidelines
  • Normally, this operand need not be specified.
  • When the value of this operand is increased, the number of synchronization point dump file input/output operations is reduced.
Operand default
When this operand is omitted, the specification of the same operand in the server common definition is assumed. When the same operand is also omitted in the server common definition, the default is 32768.
59) pd_log_sdinterval = system-log-output-volume[,interval]
Specifies the collection interval for synchronization point dumps. This operand can be specified based on the following information:
  • Volume of system log information
  • Amount of time that has elapsed since collection of the previous synchronization point dump
system-log-output-volume: ~<unsigned integer>((100-100000)) (number of log blocks)
Specifies an interval between synchronization point dumps in terms of the number of blocks of log information. A synchronization point dump is collected each time system log information equivalent to the number of log blocks specified here has been output.
interval: ~<unsigned integer>((0, 10-1440)) (minutes)
Specifies a synchronization point dump collection interval in terms of the number of minutes between synchronization point dumps.
  • If 0 is specified for the interval, HiRDB does not use a time interval for collecting synchronization point dumps.
  • If no transactions execute during the time interval since the previous synchronization point dump was collected, no synchronization point dump is collected, even if the interval specified here has elapsed.
Specification guidelines
  • This operand need not be specified if no specific amount of time is specified for restarting HiRDB.
  • The value specified for this operand affects the amount of time required to restart HiRDB.
    Specifying a small value for this operand reduces the amount of time required for database recovery during a HiRDB restart. However, because the frequency of synchronization point dumps increases, online performance might deteriorate in some cases.
    Conversely, specifying a large value for this operand increases the amount of time required for database recovery during a HiRDB restart. However, because the frequency of synchronization point dumps decreases, online performance might improve in some cases.
Tuning the specified value
The synchronization point dump collection intervals can be checked with the statistics analysis utility; the relevant information is shown under SYNC POINT GET INTERVAL in the statistical information related to system operation. Use the average of the SYNC POINT GET INTERVAL values. If the synchronization point dump collection interval is determined to be too long, decrease the specification value; conversely, if it is determined to be too short, increase the specification value.
Operand default
If this operand is omitted, the specification for the same operand in the server common definition is takes effect. If the same operand is also omitted in the server common definition, the following values are assumed:
  • system-log-output-volume: 5000
  • interval: 60
Note
  • The synchronization point dump collection interval is determined based on the volume of system log information that is output. Therefore, committing data from memory to a database takes a long time during intervals that have few updating transactions. If an error occurs at such a time, it will take longer to recover the transactions that were generated during that period. If this is a possibility, also use the interval value to set the synchronization point dump collection interval.
  • When a synchronization point dump is collected, an update page will be output from the global buffer, increasing the load on the CPU, I/O processing, and lock time. For this reason, reducing the dump interval can delay processing.