Scalable Database Server, HiRDB Version 8 Description

[Contents][Glossary][Index][Back][Next]

7.1.2 Times at which database can be recovered

A database can be recovered to its status at the following times:

Organization of this subsection
(1) Recovering to the point a backup was made
(2) Recovering to the most recent synchronization point before the failure occurred

(1) Recovering to the point a backup was made

To recover a database to its status when a backup was made, you do not need an unload file. Only the backup file is required.

(2) Recovering to the most recent synchronization point before the failure occurred

The point in time at which a transaction is completed is called a synchronization point. Synchronization point processing that validates a transaction-induced update is called a commit. Synchronization point processing that invalidates a transaction is called a rollback. Recovering a database to a transaction synchronization point is called recovering to the most recent synchronization point before the failure occurred. Because any transaction being processed when the failure occurred (a transaction that had not reached a synchronization point) is invalid, update processing performed by such transactions is not recovered. Figure 7-2 shows the transactions that are recovered.

Figure 7-2 Transactions that are recovered (with recovery to the most recent synchronization point before the failure occurred)

[Figure]

Explanation
Because processing of transactions A and B finished and they reached their synchronization points, the database is recovered to these synchronization points.
Because processing of transactions C and D was ongoing, the processing of these transactions is invalid, and is not recovered.

To recover to the most recent synchronization point before the failure occurred, in addition to backup files, you need the unload log files from which log files are output after a backup was made.