Skip Headers
Oracle® Database Backup and Recovery Reference
11g Release 1 (11.1)

Part Number B28273-01
Go to Documentation Home
Home
Go to Book List
Book List
Go to Table of Contents
Contents
Go to Index
Index
Go to Master Index
Master Index
Go to Feedback page
Contact Us

Go to previous page
Previous
Go to next page
Next
View PDF

FLASHBACK DATABASE

Purpose

Use the FLASHBACK DATABASE command to rewind the database to a target time, SCN, or log sequence number.

This command works by undoing changes made by Oracle Database to the datafiles that exist when you run the command. Flashback can fix logical failures, but not physical failures. Thus, you cannot use the command to recover from disk failures or the accidental deletion of datafiles.

FLASHBACK DATABASE is usually much faster than a RESTORE operation followed by point-in-time recovery, because the time needed to perform FLASHBACK DATABASE depends on the number of changes made to the database since the desired flashback time. On the other hand, the time needed to do a traditional point-in-time recovery from restored backups depends on the size of the database.

Flashback Database also has a number of uses in a Data Guard environment. Refer to Oracle Data Guard Concepts and Administration for details.

See Also:

Oracle Data Guard Concepts and Administration to learn about uses of Flashback Database in a Data Guard environment

Prerequisites

You can run this command from the RMAN prompt or from within a RUN command. RMAN must be connected to the target database, which must be Oracle Database 10g or later. The database must be mounted with a current control file, that is, the control file cannot be a backup or re-created. The database must run in ARCHIVELOG mode.

You cannot use FLASHBACK DATABASE to return to a point in time before the restore or re-creation of a control file. If the database control file is restored from backup or re-created, then all existing flashback log information is discarded.

The flash recovery area must be configured to enable flashback logging. Flashback logs are stored as Oracle-managed files in the flash recovery area and cannot be created if no flash recovery area is configured. You must have enabled the flashback logging before the target time for flashback by means of the SQL statement ALTER DATABASE ... FLASHBACK ON. Query V$DATABASE.FLASHBACK_ON to see whether flashback logging has been enabled.

The database must contain no online tablespaces for which flashback functionality was disabled with the SQL statement ALTER TABLESPACE ... FLASHBACK OFF.

Usage Notes

A Flashback Database operation applies to the whole database. You cannot flash back individual tablespaces. A Flashback Database operation is similar to a database point-in-time recovery (DBPITR) performed with RECOVER, but RMAN uses flashback logs to undo changes to a point before the target time or SCN. RMAN automatically restores from backup any archived redo logs that are needed and recovers the database to make it consistent. Note that RMAN never flashes back data for temporary tablespaces.

The earliest SCN that can be used for a Flashback Database operation depends on the setting of the DB_FLASHBACK_RETENTION_TARGET initialization parameter, and on the actual retention of flashback logs permitted by available disk. View the current database SCN in V$DATABASE.CURRENT_SCN.

Effect of NOLOGGING Operations on Flashback Database

When using FLASHBACK DATABASE with a target time at which a NOLOGGING operation was in progress, block corruption is likely in the database objects and datafiles affected by the NOLOGGING operation. For example, assume that you do a direct-path INSERT operation in NOLOGGING mode and that the operation runs from 9:00 to 9:15 on April 3. If you later use Flashback Database to return to 09:07 on this date, then the objects and datafiles updated by the direct-path INSERT may be left with block corruption after Flashback Database completes.

If possible, avoid using FLASHBACK DATABASE with a target time or SCN that coincides with a NOLOGGING operation. Also, perform a full or incremental backup of the affected datafiles immediately after any NOLOGGING operation to ensure recoverability to points in time after the operation. If you expect to use FLASHBACK DATABASE to return to a point in time during an operation such as a direct-path INSERT, then consider performing the operation in LOGGING mode.

See Also:

The discussion of logging_clause in Oracle Database SQL Language Reference for more information about operations that support NOLOGGING mode

Effect of Datafile Status Changes on Flashback Database

The FLASHBACK DATABASE command does not start modifying the database until it has made sure that it has all the files and resources that it needs. A Flashback Database operation should never fail due to missing datafiles, redo log files, or flashback logs.

If a datafile has changed status between the current SCN and the target SCN of the flashback, then the FLASHBACK DATABASE command behaves differently depending on the nature of the status change. Refer to Table 2-7 for details.

Table 2-7 How FLASHBACK DATABASE Responds to Datafile Status Changes

If this datafile operation occurred during the flashback window ... Then the FLASHBACK DATABASE command ...

Added

Removes the datafile record from the control file.

Dropped

Adds the datafile to the control file, but marks it as offline and does not flash it back. You can then restore and recover the datafile to the same time or SCN.

Renamed

Ignores the renaming. The datafile retains its current name.

Resized

May fail. You can take the datafile offline and then rerun the FLASHBACK DATABASE command. The datafile will not be flashed back. You can then restore and recover the datafile to the same time or SCN.

Taken offline

Ignores the operation. The datafile retains its current online status.

Brought online

Ignores the operation. The datafile retains its current offline status.

Made read-only or read/write

Changes the status of the datafile in the control file.


Tablespaces with Flashback Logging Disabled

It is possible for the ALTER TABLESPACE ... FLASHBACK OFF statement to have been executed for some tablespaces. If FLASHBACK DATABASE has insufficient flashback data to rewind a tablespace to the target SCN, then RMAN issues an error and does not modify the database. Whenever FLASHBACK DATABASE fails or is interrupted, the database is left mounted.

In this scenario, query V$TABLESPACE to determine which tablespaces have flashback logging disabled. You have the following options:

State of the Database After Flashback Database

After running FLASHBACK DATABASE, the database may not be left at the SCN most immediately before the target time. Events other than transactions can cause the database SCN to be updated. If you use the FLASHBACK DATABASE TO form of the command, and if a transaction is associated with the target SCN, then after the flashback the database will include all changes up to and including this transaction. Otherwise, all changes up to but not including this transaction will be included in the datafiles, whether you use the FLASHBACK DATABASE TO or FLASHBACK DATABASE TO BEFORE form of the command. Changes after the specified target SCN are never applied as a result of a FLASHBACK DATBASE.

After FLASHBACK DATABASE completes, you may want to open the database read-only and run queries to ensure that you achieved the intended result. If you are not satisfied, then you can use RECOVER DATABASE to recover the database to its state when you started the flashback. You can then rerun FLASHBACK DATABASE.

If you are satisfied with the results of the flashback, then you can OPEN RESETLOGS to abandon all changes after the target time. Alternatively, you can use Data Pump to export lost data, use RECOVER DATABASE to return the database to its state before the flashback operation, and then use Data Pump to reimport the lost data.

Syntax

flashback::=

Description of flashback.gif follows
Description of the illustration flashback.gif

(deviceSpecifier::=)

Semantics

Syntax Element Description
DEVICE TYPE deviceSpecifier Allocates automatic channels for the specified device type only. For example, if you configure automatic disk and tape channels, and issue FLASHBACK...DEVICE TYPE DISK, then RMAN allocates only disk channels. RMAN may need to restore redo logs from backup during the flashback database process. Changes between the last flashback log and the target time must be re-created based on the archived redo log. If no automatic channels are allocated for tape and a needed redo log is on tape, then the FLASHBACK DATABASE operation fails.

See Also: deviceSpecifier

TO BEFORE SCN integer Returns the database to its state just before the specified SCN. Any changes at an SCN lower than that specified are applied, but if there is a change associated with the specified SCN it is not applied. By default, the provided SCN resolves to the current or ancestor incarnation. You can override the default by using the RESET DATABASE INCARNATION command.

Query OLDEST_FLASHBACK_SCN in V$FLASHBACK_DATABASE_LOG to see the approximate lowest SCN to which you can flash back.

TO BEFORE SEQUENCE integer [THREAD integer] Specifies a redo log sequence number and thread as an upper limit. RMAN applies changes up to (but not including) the last change in the log with the specified sequence and thread number.
TO BEFORE RESETLOGS Returns the database to its state including all changes up to the SCN of the most recent OPEN RESETLOGS.

Note: FLASHBACK DATABASE can only return the database to a point before the most recent OPEN RESETLOGS operation if your database has been upgraded to Oracle Database 10g Release 2 or later.

TO BEFORE TIME 'date_string' Returns the database to its state including all changes up to but not including changes at the specified time.

Query OLDEST_FLASHBACK_TIME in V$FLASHBACK_DATABASE_LOG to see the approximate lowest time to which you can flash back.

TO SCN integer Returns the database to the point up to (and including) the specified SCN. By default, the provided SCN resolves to the current or ancestor incarnation. You can override the default by using the RMAN RESET DATABASE command to set the recovery target incarnation.

Query OLDEST_FLASHBACK_SCN in V$FLASHBACK_DATABASE_LOG to see the approximate lowest SCN to which you can flash back.

TO SEQUENCE integer THREAD integer Specifies a redo log sequence number and thread as an upper limit. RMAN applies changes up to (and including) the last change in the log with the specified sequence and thread number.
TO RESTORE POINT restore_point_name Returns the database to the SCN associated with the specified restore point. This can be an ordinary restore point or a guaranteed restore point.
TO TIME 'date_string' Returns the database to its state at the specified time. You can use any SQL DATE expressions to convert the time to the current format, for example, FLASHBACK DATABASE TO TIME 'SYSDATE-7'.

Query OLDEST_FLASHBACK_TIME in V$FLASHBACK_DATABASE_LOG to see the approximate lowest time to which you can flash back.


Examples

Example 2-79 FLASHBACK DATABASE to a Specific SCN

Assume that you inserted corrupted database in many tables at 5:00 p.m. on February 14. The following statement queries the earliest SCN in the flashback window:

SQL> SELECT OLDEST_FLASHBACK_SCN, OLDEST_FLASHBACK_TIME
  2  FROM   V$FLASHBACK_DATABASE_LOG;

OLDEST_FLASHBACK_SCN OLDEST_FLASHBACK
-------------------- ----------------
              411010 2007/02/14 16:49

You start an RMAN session and enter commands as follows (sample output for the FLASHBACK DATABASE is included):

RMAN> CONNECT TARGET /
RMAN> CONNECT CATALOG rman/password@catdb
RMAN> SHUTDOWN IMMEDIATE
RMAN> STARTUP MOUNT
RMAN> FLASHBACK DATABASE TO SCN 411010;
 
Starting flashback at 15-FEB-07
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=104 device type=DISK
 
 
starting media recovery
media recovery complete, elapsed time: 00:00:07
 
Finished flashback at 15-FEB-07
 
RMAN> ALTER DATABASE OPEN RESETLOGS;

Example 2-80 FLASHBACK DATABASE to a Restore Point

Assume that you are preparing to load a massive number of updates to the database. You create a guaranteed restore point before the updates:

SQL> CREATE RESTORE POINT before_update GUARANTEE FLASHBACK DATABASE;

The update fails, leaving the database with extensive corrupted data. You start an RMAN session and list the guaranteed restore points:

RMAN> CONNECT TARGET /
RMAN> CONNECT CATALOG rman/password@catdb
RMAN> LIST RESTORE POINT ALL;

SCN              RSP Time  Type       Time      Name
---------------- --------- ---------- --------- ----
412742                     GUARANTEED 15-FEB-07 BEFORE_UPDATE

You mount the database, flash back the database to the restore point (sample output included), and then open the database with the RESETLOGS option:

RMAN> SHUTDOWN IMMEDIATE
RMAN> STARTUP MOUNT
RMAN> FLASHBACK DATABASE TO RESTORE POINT 'BEFORE_UPDATE';
 
Starting flashback at 15-FEB-07
allocated channel: ORA_DISK_1
channel ORA_DISK_1: SID=104 device type=DISK
 
 
starting media recovery
 
archived log for thread 1 with sequence 34 is already on disk as file /disk2/oracle/oradata/prod/arch/archive1_34_614598462.dbf
media recovery complete, elapsed time: 00:00:01
Finished flashback at 15-FEB-07
 
RMAN> ALTER DATABASE OPEN RESETLOGS;