In cases of transient system failure, system recovery occurs automatically either the next time any user attempts to login to a location in which a transaction was abnormally terminated or the next time a SET LOCATION command is issued against an affected location. For any transaction that was abnormally terminated before it completed, the before image stored in the location log files is used to automatically rollback all uncommitted transactions and restore the data tables involved to a logically consistent state.
This transient system failure recovery feature (the same functionality provided by backward logging in previous versions of the XDB Server) is permanently defaulted to on, with no user intervention required.
The average length of time necessary to bring all location data objects to a logically consistent state is affected by the value of the Checkpoint Frequency configuration parameter. This parameter allows the user to adjust the time between automatic logging system checkpoints. See the section Server Configuration for information about setting the Checkpoint Frequency parameter.
When the logging system issues a checkpoint, the contents of the entire log buffer cache are flushed to disk, that is, written to table files.