2.3.6 Operands related to full recovery processing
- 34) pd_max_recover_process = concurrently-executable-full-recovery-processes-count
<unsigned integer>((1-10))<<3>>- Specifies the number of processes to be recovered during full recovery processing. For a HiRDB/Parallel Server, this operand specifies the number of processes to be recovered per server (dictionary server or back-end server).
- Condition
- You need to use the raw I/O facility.
- Specification guidelines
- There are three or more logical volumes for which an RDAREA is defined (per server): 3
- There are fewer than three logical volumes for which an RDAREA is defined (per server): Number of logical volumes
- Increasing the value of this operand increases the input/output concurrency during full recovery processing, and thus can shorten the recovery time. However, because a number of processes equaling value-of-this-operand
server-count are started, determine a value by taking the aforementioned specification value guideline and HiRDB resources into consideration.
- 35) pd_redo_allpage_put = Y | N
- Specifies whether to output to a database the pages that were updated after a synchronization point during full recovery processing.
- Y:
- During full recovery processing, all pages updated after a synchronization point are output.
- N:
- During full recovery processing, only those pages that were not output to the database when an error occurred are output.
- Specification guidelines
- Normally, there is no need to specify this operand.
- When a database is mirrored using LVM (Logical Volume Manager) or the device driver mirroring facility, the original and duplicate of the mirror may not match each other if an OS or machine error occurs or if a unit error results in system switchover. In this case, to synchronize the original and duplicate volumes:
- First use LVM or the device driver facility to synchronize the original and duplicate volumes, and then restart HiRDB.
- If you cannot use the above method to synchronize the original and duplicate volumes, or if the synchronization operation takes too long to satisfy the system requirements such as the system switchover time, specify Y for this operand. During full recovery processing at the restart of HiRDB following an error, all pages that were updated after a synchronization point are output to a database. In this process, HiRDB synchronizes the original and duplicate volumes, eliminating the mismatch between the original and duplicate mirrors.
- For the action to take when a mismatch occurs between the original and duplicate mirrors, see the HiRDB Version 8 System Operation Guide.
- Notes
- When Y is specified for this operand, there is overhead for outputting all pages to the database during full recovery processing.
- When this operand is specified, the message KFPH24004-I is output when the full recovery processing is finished. No message is output if this operand is omitted.