![[Contents]](FIGURE/CONTENT.GIF)
Scalable Database Server, HiRDB Version 8 Description
![[Glossary]](FIGURE/GLOSS.GIF)
![[Index]](FIGURE/INDEX.GIF)
![[Back]](FIGURE/FRONT.GIF)
![[Next]](FIGURE/AFTER.GIF)
If system logs are the only system files used to recover HiRDB when it has terminated abnormally, recovery processing must employ every system log from the point in time that HiRDB was last started. For this reason, recovering the system from system logs alone may take a significant amount of time. To mitigate this issue, points (called synchronization points) are established at fixed time intervals while HiRDB is running, and management information (synchronization point dumps) needed for recovery operations is saved at these points. This eliminates the need for system logs prior to the last synchronization point, which shortens the system recovery time.
At each synchronization point, HiRDB incorporates into the database the details of all the database updates that have been made since the last synchronization point or since HiRDB was started. The HiRDB administrator should create synchronization point dump files as a safeguard against failures.
HiRDB operates with logical units of synchronization point dump files called file groups. A file group consists of either one or two synchronization point dump files. A configuration of two synchronization point dump files is called a duplexed synchronization point dump file. The respective synchronization point dump files are distinguished by referring to them as file A and file B. When synchronization point dump files are duplexed, HiRDB collects and saves the same synchronization point dump information into both files. This enhances system reliability because, even if an error occurs in one of the files, the other file is still available. Figure 6-3 shows the organization of synchronization point dump files.
Figure 6-3 Organization of synchronization point dump files
![[Figure]](FIGURE/ZU060030.GIF)
If the most recent synchronization point dump file that has been generated cannot be read because an error has occurred in the file, HiRDB attempts to read the file that was created one generation earlier. If HiRDB cannot read the file that was created one generation earlier, it attempts to read the file that was created one generation previous to that one. In this way, HiRDB continues attempting to read previous generations of files if the read attempt fails.
The number of guaranteed valid generations refers to the number of previous generations of synchronization point dump files for which overwriting is prohibited. For example, if the number of guaranteed valid generations is 2, two generations of synchronization point dump files (the most recent generation and one generation previous to that one) cannot be overwritten. Thus, increasing the number of guaranteed valid generations enhances system reliability because, even if an error occurs in a synchronization point dump file, synchronization point dump files of the number set in the number of guaranteed valid generations will always be available.
Note that the number of synchronization point dump files you need is the number of guaranteed valid generations + 1.
The pdloginit command is used to create synchronization point dump files. Additionally, you specify the following HiRDB system definition operands to create an environment in which the synchronization point dump files can be used:
For details about designing and creating synchronization point dump files, see the HiRDB Version 8 Installation and Design Guide. For details about using synchronization point dump files, see the HiRDB Version 8 System Operation Guide.
All Rights Reserved. Copyright (C) 2007, Hitachi, Ltd.