Appendix A.2 Synchronization point dump files
(1) The KFPS02179-I message is issued repeatedly
- Question
- Why is the KFPS02179-I message being issued repeatedly (this message says that the time for collecting a synchronization point dump has been reached, but collection of the dump was skipped because the previous synchronization point dump processing had not been completed)?
- Answer
- This message is issued when HiRDB is waiting for commitment of an updating transaction that was being executed when a synchronization point dump was collected (i.e., HiRDB is waiting for completion of validation). There should be no problem if there is a sufficient number of system log files with sufficient space in them. However, if the unit terminates abnormally in such status, the restart time for HiRDB or the unit may be prolonged.
- If there are enough generations and there is enough space in the system log file, HiRDB or the unit may terminate abnormally due to a shortage of overwrite-enabled system log files (the KFPS01256-E message is issued). To avoid such problems, reevaluate the following:
- When it takes a long time for the transaction to commit after updating a database or data dictionary
- Evaluate the following items:
- Would it be possible to issue the commit statement during the processing?
- If the transaction retrieves a large amount of data after updating, would it be possible to reduce the time from the first update processing in the transaction to the commit point by changing the sequence of the update processing and the retrieval processing?
- When there is insufficient space in the system log file or too few generations
- Evaluate the following items:
- It is recommended that you provide the amount of file space necessary to store system logs that will be output during operation of HiRDB. If this is not feasible, try to allocate at least the size recommended in the HiRDB Version 8 Installation and Design Guide.
- Do not collect a database update log for a transaction that updates a large amount of data. Whether or not a database update log is to be collected is specified in the PDDBLOG operand in the client environment definition. If a database update log is not collected and an error occurs during UAP execution, the database can be restored only to the point where its backup was made.
(2) Number of guaranteed valid generations
- Question
- What is meant by the number of guaranteed valid generations of the synchronization point dump file?
- Answer
- Information such as a read operation's starting point in the system log file is acquired in the synchronization point dump file; this information is useful in the event a restart of HiRDB becomes necessary. Thus, the portion of the system log file beginning at the location indicated in the synchronization point dump file is write-protected because it might be needed for a restart.
- The number of guaranteed valid generations is the number of generations of synchronization point dump files that are used to write-protect the system log file. In other words, if the number of guaranteed valid generations is 1, only the system log file corresponding to the most recent synchronization point dump file is write-protected. If the number of guaranteed valid generations is 2, the system log file corresponding to the most recent synchronization point dump file and the one immediately preceding it are write-protected. Therefore, if the number of guaranteed valid generations is 2, the number of system log files that are write-protected may increase.
- The number of synchronization point dump file guaranteed valid generations is specified in the pd_spd_assurance_count operand.