DB_LOST_WRITE_PROTECT
Posted: Tue Oct 09, 2012 8:41 pm
DB_LOST_WRITE_PROTECT is a new initialization parameter seen starting Oracle 11g Release 1. This enables or disables lost write detection. A data block lost write occurs when an I/O subsystem acknowledges the completion of the block write, while in fact the write did not occur in the persistent storage.
This lost-write protection feature works best in a Data Guard environment where you set DB_LOST_WRITE_PROTECT parameter in both primary and standby databases.
DB_LOST_WRITE_PROTECT can be set to NONE, TYPICAL, or FULL.
Below is the description of its value:
1. When the parameter is set to TYPICAL on the primary database, the instance logs buffer cache reads for read-write tablespaces in the redo log, which is necessary for detection of lost writes.
2. When the parameter is set to FULL on the primary database, the instance logs reads for read-only tablespaces as well as read-write tablespaces.
3. When the parameter is set to TYPICAL or FULL on the standby database or on the primary database during media recovery, the instance performs lost write detection.
4. When the parameter is set to NONE on either the primary database or the standby database, no lost write detection functionality is enabled.
When the redo is being applied on the standby database, the SCN numbers are compared between the two systems. If a lower number is found on the primary database, an ORA-752 error is recorded into the alert log. If a lower number is found on the standby database, an ORA-600 [3020] error is recorded.
Query the V$DATABASE_BLOCK_CORRUPTION view.
The best solution if a lost write is detected on the primary system is to fail over to your standby and recreate the primary. If it is detected on the standby system, you should recreate the standby system, or at least the affected files.
This lost-write protection feature works best in a Data Guard environment where you set DB_LOST_WRITE_PROTECT parameter in both primary and standby databases.
DB_LOST_WRITE_PROTECT can be set to NONE, TYPICAL, or FULL.
Below is the description of its value:
1. When the parameter is set to TYPICAL on the primary database, the instance logs buffer cache reads for read-write tablespaces in the redo log, which is necessary for detection of lost writes.
2. When the parameter is set to FULL on the primary database, the instance logs reads for read-only tablespaces as well as read-write tablespaces.
3. When the parameter is set to TYPICAL or FULL on the standby database or on the primary database during media recovery, the instance performs lost write detection.
4. When the parameter is set to NONE on either the primary database or the standby database, no lost write detection functionality is enabled.
When the redo is being applied on the standby database, the SCN numbers are compared between the two systems. If a lower number is found on the primary database, an ORA-752 error is recorded into the alert log. If a lower number is found on the standby database, an ORA-600 [3020] error is recorded.
Query the V$DATABASE_BLOCK_CORRUPTION view.
The best solution if a lost write is detected on the primary system is to fail over to your standby and recreate the primary. If it is detected on the standby system, you should recreate the standby system, or at least the affected files.