2.11.1 Recovery flow based on a restart
If the HADB server terminated abnormally, restarting HADB causes the HADB server to automatically restart using the status information. Once restarted, HADB uses the system logs to recover the database and restores it to the state it was in before the HADB server terminated abnormally.
- Organization of this subsection
(1) Status information
To guard against abnormal termination, the HADB server outputs the information necessary for automatic restarting. This information is called the status information.
Status information is output whenever the status of the HADB server changes, and it is stored in the status file. The following types of information are output as status information:
-
HADB server termination mode (normal termination, abnormal termination, and so on)
-
HADB server operational status (start processing, running, termination processing, and so on)
The status file is stored under the DB directory ($DBDIR/ADBSYS/ADBSTS).
(2) System log
When a database is updated, its update history information is output. This information is called the system log. Files in which the system logs are stored is called system log files.
System logs are used for restarting HADB after an abnormal termination, and for the recovery processing after a rollback has occurred.
(a) System log file creation location
The system log files are stored under the DB directory ($DBDIR/ADBSYS/ADBSLG).
(b) System log file configuration
System log files actually consist of two different types of files, a master log file and user log files. The following figure shows the configuration of the system log files.
|
|
![[Figure]](GRAPHICS/F0000045.GIF)
-
Master log file
The master log file stores the system control-related history information (event log) such as the start and end of transactions. It also manages user log file groups.
When the HADB server starts, a single master log file is automatically created. The file name is MSTLOG.
-
User log file
A user log file stores database-related operation history (user logs).
A separate user log file is allocated for each transaction that updates the database and for each real thread.
When the HADB server starts, a user log file is created based on settings specified by the user. The file name begins with USRLOG.
If the specifications related to user log files are changed, the system log files are re-created when the HADB server is started.