12 Using RMAN to Back Up and Restore Files

12.1.1 Interchangeability of Backups in an Oracle Data Guard Environment

RMAN commands use the recovery catalog metadata to behave transparently across different physical databases in the Oracle Data Guard environment. For example, you can back up a tablespace on a physical standby database and restore and recover it on the primary database. Similarly, you can back up a tablespace on a primary database and restore and recover it on a physical standby database.

Backups of standby control files and nonstandby control files are interchangeable. For example, you can restore a standby control file on a primary database and a primary control file on a physical standby database. This interchangeability means that you can offload control file backups to one database in an Oracle Data Guard environment. RMAN automatically updates the filenames for database files during restore and recovery at the databases.

12.1.2 Association of Backups in an Oracle Data Guard Environment

The recovery catalog tracks the files in the Oracle Data Guard environment by associating every database file or backup file with a DB_UNIQUE_NAME. The database that creates a file is associated with the file. For example, if RMAN backs up the database with the unique name of standby1, then standby1 is associated with this backup. A backup remains associated with the database that created it unless you use the CHANGE ... RESET DB_UNIQUE_NAME to associate the backup with a different database.

12.1.3 Accessibility of Backups in an Oracle Data Guard Environment

The accessibility of a backup is different from its association. In an Oracle Data Guard environment, the recovery catalog considers disk backups as accessible only to the database with which it is associated, whereas tape backups created on one database are accessible to all databases. If a backup file is not associated with any database, then the row describing it in the recovery catalog view shows null for the SITE_KEY column. By default, RMAN associates files whose SITE_KEY is null with the target database.

RMAN commands such as BACKUP, RESTORE, and CROSSCHECK work on any accessible backup. For example, for a RECOVER COPY operation, RMAN considers only image copies that are associated with the database as eligible to be recovered. RMAN considers the incremental backups on disk and tape as eligible to recover the image copies. In a database recovery, RMAN considers only the disk backups associated with the database and all files on tape as eligible to be restored.

To illustrate the differences in backup accessibility, assume that databases prod and standby1 reside on different hosts. RMAN backs up data file 1 on prod to /prmhost/disk1/df1.dbf on the production host and also to tape. RMAN backs up data file 1 on standby1 to /sbyhost/disk2/df1.dbf on the standby host and also to tape. If RMAN is connected to database prod, then you cannot use RMAN commands to perform operations with the /sbyhost/disk2/df1.dbf backup located on the standby host. However, RMAN does consider the tape backup made on standby1 as eligible to be restored.

12.3.1 Oracle Database Configurations on Primary and Standby Databases

The following Oracle Database configurations are recommended on every primary and standby database in the Oracle Data Guard environment:

• Configure a fast recovery area for each database (the recovery area is local to a database).

  The fast recovery area is a single storage location on a file system or Oracle Automatic Storage Management (Oracle ASM) disk group where all files needed for recovery reside. These files include the control file, archived logs, online redo logs, flashback logs, and RMAN backups. As new backups and archived logs are created in the fast recovery area, older files (which are either outside of the retention period, or have been backed up to tertiary storage) are automatically deleted to make room for them. In addition, notifications can be set up to alert the DBA when space consumption in the fast recovery area is nearing its predefined limit. The DBA can then take action, such as increasing the recovery area space limit, adding disk hardware, or decreasing the retention period.

  Set the following initialization parameters to configure the fast recovery area:

  DB_RECOVERY_FILE_DEST = <mount point or Oracle ASM Disk Group>
  DB_RECOVERY_FILE_DEST_SIZE = <disk space quota>
  

  See Also:

  Oracle Database Backup and Recovery User's Guide for more information about configuring a fast recovery area

• Use a server parameter file (SPFILE) so that it can be backed up to save instance parameters in backups.

• Enable Flashback Database on primary and standby databases.

  When Flashback Database is enabled, Oracle Database maintains flashback logs in the fast recovery area. These logs can be used to roll the database back to an earlier point in time, without requiring a complete restore.

  See Also:

  Oracle Database Backup and Recovery User's Guide for more information about enabling Flashback Database

12.3.2 RMAN Configurations at the Primary Database

To simplify ongoing use of RMAN, you can set a number of persistent configuration settings for each database in the Oracle Data Guard environment. These settings control many aspects of RMAN behavior. For example, you can configure the backup retention policy, default destinations for backups to tape or disk, default backup device type, and so on. You can use the CONFIGURE command to set and change RMAN configurations. The following RMAN configurations are recommended at the primary database:

1. Connect RMAN to the primary database and recovery catalog.

2. Configure the retention policy for the database as n days:

   CONFIGURE RETENTION POLICY TO RECOVERY WINDOW OF <n> DAYS;
   

   This configuration lets you keep the backups necessary to perform database recovery to any point in time within the specified number of days.

   Use the DELETE OBSOLETE command to delete any backups that are not required (per the retention policy in place) to perform recovery within the specified number of days.

3. Specify when archived logs can be deleted with the CONFIGURE ARCHIVELOG DELETION POLICY command. For example, if you want to delete logs after ensuring that they shipped to all destinations, use the following configuration:

   CONFIGURE ARCHIVELOG DELETION POLICY TO SHIPPED TO ALL STANDBY;
   

   If you want to delete logs after ensuring that they were applied on all standby destinations, use the following configuration:

   CONFIGURE ARCHIVELOG DELETION POLICY TO APPLIED ON ALL STANDBY;
   

4. Configure the connect string for the primary database and all standby databases, so that RMAN can connect remotely and perform resynchronization when the RESYNC CATALOG FROM DB_UNIQUE_NAME command is used. When you connect to the target instance, you must provide a net service name. This requirement applies even if the other database instance from where the resynchronization is done is on the local host. The target and remote instances must use the same SYSDBA (or SYSBACKUP) password, which means that both instances must already have password files. You can create the password file with a single password so you can start all the database instances with that password file. For example, if the TNS alias to connect to a standby in Boston is boston_conn_str, you can use the following command to configure the connect identifier for the BOSTON database site:

   CONFIGURE DB_UNIQUE_NAME BOSTON CONNECT IDENTIFIER 'boston_conn_str';
   

   Note that the 'boston_conn_str' does not include a username and password. It contains only the Oracle Net service name that can be used from any database site to connect to the BOSTON database site.

   After connect identifiers are configured for all standby databases, you can verify the list of standbys by using the LIST DB_UNIQUE_NAME OF DATABASE command.

12.3.3 RMAN Configurations at a Standby Database Where Backups are Performed

The following RMAN configurations are recommended at a standby database where backups are done:

1. Connect RMAN to the standby database (where backups are performed) as target, and to the recovery catalog.

2. Enable automatic backup of the control file and the server parameter file:

   CONFIGURE CONTROLFILE AUTOBACKUP ON;
   

3. Skip backing up data files for which there already exists a valid backup with the same checkpoint:

   CONFIGURE BACKUP OPTIMIZATION ON;
   

4. Configure the tape channels to create backups as required by media management software:

   CONFIGURE CHANNEL DEVICE TYPE SBT PARMS '<channel parameters>';
   

5. Because the archived logs are backed up at the standby database, Oracle recommends that you configure the BACKED UP option for the log deletion policy:

   CONFIGURE ARCHIVELOG DELETION POLICY BACKED UP n TIMES TO DEVICE TYPE SBT;
   

12.3.4 RMAN Configurations at a Standby Where Backups Are Not Performed

The following RMAN configurations are recommended at a standby database where backups are not done:

1. Connect RMAN to the standby database as target, and to the recovery catalog.

2. Enable automatic deletion of archived logs once they are applied at the standby database (this is also applicable to all terminal databases when the cascading or far sync instance features are in use):

   CONFIGURE ARCHIVELOG DELETION POLICY TO APPLIED ON ALL STANDBY;
   

12.4.1 Using Disk as Cache for Tape Backups

The fast recovery area on the standby database can serve as a disk cache for tape backup. Disk is used as the primary storage for backups, with tape providing long term, archival storage. Incremental tape backups are taken daily and full tape backups are taken weekly. The commands used to perform these backups are described in the following sections.

RESYNC CATALOG FROM DB_UNIQUE_NAME ALL;
RECOVER COPY OF DATABASE WITH TAG 'OSS';
BACKUP DEVICE TYPE DISK INCREMENTAL LEVEL 1 FOR RECOVER OF COPY WITH TAG 'OSS' DATABASE;
BACKUP DEVICE TYPE SBT ARCHIVELOG ALL;
BACKUP BACKUPSET ALL;
DELETE ARCHIVELOG ALL;

BACKUP RECOVERY FILES;

12.4.2 Performing Backups Directly to Tape

Oracle's Media Management Layer (MML) API lets third-party vendors build a media manager, software that works with RMAN and the vendor's hardware to allow backups to sequential media devices such as tape drives. A media manager handles loading, unloading, and labeling of sequential media such as tapes. You must install Oracle Secure Backup or third-party media management software to use RMAN with sequential media devices.

Take the following steps to perform backups directly to tape, by default:

1. Connect RMAN to the standby database (as the target database) and recovery catalog.

2. Execute the CONFIGURE command as follows:

   CONFIGURE DEFAULT DEVICE TYPE TO SBT;
   

In this scenario, full backups are taken weekly, with incremental backups taken daily on the standby database.

12.7.1 Changing Metadata in the Recovery Catalog

You can use the RMAN CHANGE command with various operands to change metadata in the recovery catalog, as described in the following sections.

Changing File Association from One Standby Database to Another

Use the CHANGE command with the RESET DB_UNIQUE_NAME option to alter the association of files from one database to another within an Oracle Data Guard environment. The CHANGE command is useful when disk backups or archived logs are transferred from one database to another and you want to use them on the database to which they were transferred. The CHANGE command can also change the association of a file from one database to another database, without having to directly connect to either database using the FOR DB_UNIQUE_NAME and RESET DB_UNIQUE_NAME TO options.

Changing DB_UNIQUE_NAME for a Database

If the value of the DB_UNIQUE_NAME initialization parameter changes for a database, the same change must be made in the Oracle Data Guard environment. The RMAN recovery catalog, after connecting to that database instance, will know both the old and new value for DB_UNIQUE_NAME. To merge the information for the old and new values within the recovery catalog schema, you must use the RMAN CHANGE DB_UNIQUE_NAME command. If the value of the DB_UNIQUE_NAME initialization parameter changes for a database, the same change must be made in RMAN so that it is aware of the new DB_UNIQUE_NAME. For example, perform the following steps to change the database with DB_UNIQUE_NAME of BOSTON_A to BOSTON_B:

1. In the initialization parameter file or SQL, change the DB_UNIQUE_NAME initialization parameter from BOSTON_A to BOSTON_B.

2. In RMAN, connect to any database in the Oracle Data Guard environment as target database and connect to the recovery catalog. Then execute the CHANGE command:

   CHANGE DB_UNIQUE_NAME FROM BOSTON_A TO BOSTON_B;
   

Making Backups Unavailable or Removing Their Metadata

Use CHANGE command options such as AVAILABLE, UNAVAILABLE, KEEP, and UNCATALOG to make backups available or unavailable for restore and recovery purposes, and to keep or remove their metadata.

12.7.2 Deleting Archived Logs or Backups

Use the DELETE command to delete backup sets, image copies, archived logs, or proxy copies. To delete only files that are associated with a specific database, you must use the FOR DB_UNIQUE_NAME option with the DELETE command.

File metadata is deleted for all successfully deleted files associated with the current target database (or for files that are not associated with any known database). If a file could not be successfully deleted, you can use the FORCE option to remove the file's metadata.

When a file associated with another database is deleted successfully, its metadata in the recovery catalog is also deleted. Any files that are associated with other databases, and that could not be successfully deleted, are listed at the completion of the DELETE command, along with instructions for you to perform the same operation at the database with which the files are associated (files are grouped by database). Note that the FORCE option cannot be used to override this behavior. If you are certain that deleting the metadata for the non-deletable files will not cause problems, you can use the CHANGE RESET DB_UNIQUE_NAME command to change the metadata for association of files with the database and use the DELETE command with the FORCE option to delete the metadata for the file.

12.7.3 Validating Recovery Catalog Metadata

Use the CROSSCHECK command to validate and update file status in the recovery catalog schema. To validate files associated with a specific database, use the FOR DB_UNIQUE_NAME option with the CROSSCHECK command.

Metadata for all files associated with the current target database (or for any files that are not associated with any database), will be marked AVAILABLE or EXPIRED according to the results of the CROSSCHECK operation.

If a file associated with another database is successfully inspected, its metadata in the recovery catalog is also changed to AVAILABLE. Any files that are associated with other databases, and that could not be inspected successfully, are listed at the completion of the CROSSCHECK command, along with instructions for you to perform the same operation at the database with which the files are associated (files are grouped by site). If you are certain of the configuration and still want to change status metadata for unavailable files, you can use the CHANGE RESET DB_UNIQUE_NAME command to change metadata for association of files with the database and execute the CROSSCHECK command to update status metadata to EXPIRED.

12.8.1 Recovery from Loss of Files on the Primary or Standby Database

As of Oracle Database 12c Release 1 (12.1), you can restore and recover files over the network by connecting to a physical standby database that contains the required files. This can be useful when you want to restore lost data files, control files, or tablespaces on a primary database using the corresponding files on the physical standby database. You can also use the same process to restore files on a physical standby database by using the primary database.

For an example of how to restore and recover files by connecting over the network, see Oracle Database Backup and Recovery User's Guide.

12.8.2 Recovery from Loss of Online Redo Log Files

If all online log members for the current ACTIVE group or for an inactive group which has not yet been archived are lost, then you must fail over to the standby database. Refer to Chapter 9 for the failover procedure.

For information about how to recover from the loss of online redo log files in other circumstances, see Oracle Database Backup and Recovery User's Guide.

12.8.3 Incomplete Recovery of the Primary Database

Incomplete recovery of the primary database is normally done in cases such as when the database is logically corrupted (by a user or an application) or when a tablespace or data file was accidentally dropped from database.

Depending on the current database checkpoint SCN on the standby database instances, you can use one of the following procedures to perform incomplete recovery of the primary database. All the procedures are in order of preference, starting with the one that is the least time consuming.

Using Flashback Database Using Flashback Database is the recommended procedure when the Flashback Database feature is enabled on the primary database, none of the database files are lost, and the point-in-time recovery is greater than the oldest flashback SCN or the oldest flashback time. See Section 15.3 for the procedure to use Flashback Database to do point-in-time recovery.

Using the standby database instance This is the recommended procedure when the standby database is behind the desired incomplete recovery time, and Flashback Database is not enabled on the primary or standby databases:

1. Recover the standby database to the desired point in time. Be sure to stop the managed redo process (MRP) before issuing the following command:

   RECOVER DATABASE UNTIL TIME 'time';
   

   Alternatively, incomplete recovery time can be specified using the SCN or log sequence number:

   RECOVER DATABASE UNTIL SCN incomplete recovery SCN';
   RECOVER DATABASE UNTIL LOGSEQ incomplete recovery log sequence number THREAD thread number;
   

2. Open the standby database in read-only mode to verify the state of database.

   If the state is not what is desired, use the LogMiner utility to look at the archived redo log files to find the right target time or SCN for incomplete recovery. Alternatively, you can start by recovering the standby database to a point that you know is before the target time, and then open the database in read-only mode to examine the state of the data. Repeat this process until the state of the database is verified to be correct. Note that if you recover the database too far (that is, past the SCN where the error occurred) you cannot return it to an earlier SCN.

3. Activate the standby database using the SQL ALTER DATABASE ACTIVATE STANDBY DATABASE statement. This converts the standby database to a primary database, creates a new resetlogs branch, and opens the database. See Section 10.4 to learn how the standby database reacts to the new reset logs branch.

Using the primary database instance If all of the standby database instances have already been recovered past the desired point in time and Flashback Database is not enabled on the primary or standby database, then this is your only option.

Use the following procedure to perform incomplete recovery on the primary database:

1. Use LogMiner or another means to identify the time or SCN at which all the data in the database is known to be good.

2. Using the time or SCN, issue the following RMAN commands to do incomplete database recovery and open the database with the RESETLOGS option (after connecting to catalog database and primary instance that is in MOUNT state):

   RUN 
   {
   SET UNTIL TIME 'time';
   RESTORE DATABASE;
   RECOVER DATABASE;
   }
   ALTER DATABASE OPEN RESETLOGS;
   

After this process, all standby database instances must be reestablished in the Oracle Data Guard configuration.

12.8.4 Actions Needed on Standby After TSPITR or Tablespace Plugin at Primary

After an RMAN tablespace point-in-time recovery (TSPITR) is performed at the primary, the recovered data files have a new system change number (SCN), and are therefore treated like new data files at the primary. These data files cannot be automatically created at the standby.

Likewise, when a new plugged in tablespace is added to the primary database, the data files are treated like new data files at the primary.

The managed redo process (MRP) at the standby stops when the Redo Apply process encounters creation of these new files. The required new data files must be copied and restored to the standby. You can do this using either backups or a direct copy from the primary. For example, to copy all files that belong to a tablespace that has undergone an RMAN TSPITR, you can use the following RMAN command:

RMAN> RESTORE TABLESPACE <tbs_name1, tbs_name2> FROM SERVICE <tnsalias-of-primary>

Note that the number of disk channels allocated will be as per RMAN configurations. So, if CONFIGURE DEVICE TYPE DISK PARALLELISM 4 is executed, then 4 disk channels will be used to pull the files from the primary database.

When the new data files are available at the standby, MRP should be restarted to continue applying the logs.

12.9.1 Standby Databases Too Geographically Distant to Share Backups

If the standby databases are far apart from one another, the backups taken on them may not be easily accessible by the primary system or other standby systems. Perform a complete backup of the database on all systems to perform recovery operations. The fast recovery area can reside locally on the primary and standby systems (that is, the fast recovery area does not have to be the same for the primary and standby databases).

In this scenario, you can still use the general strategies described in Section 12.8, with the following exceptions:

• Backup files created by RMAN must be tagged with the local system name, and with RESTORE operations that tag must be used to restrict RMAN from selecting backups taken on the same host. In other words, the BACKUP command must use the TAG system name option when creating backups; the RESTORE command must use the FROM TAG system name option; and the RECOVER command must use the FROM TAG system name ARCHIVELOG TAG system name option.

• Disaster recovery of the standby site:

  1. Start the standby instance in the NOMOUNT state using the same parameter files with which the standby was operating earlier.

  2. Create a standby control file on the primary instance using the SQL ALTER DATABASE CREATE STANDBY CONTROLFILE AS filename statement, and use the created control file to mount the standby instance.

  3. Issue the following RMAN commands to restore and recover the database files:

     RESTORE DATABASE FROM TAG 'system name';
     RECOVER DATABASE FROM TAG 'system name' ARCHIVELOG TAG 'system name';
     

  4. Restart Redo Apply.

The standby instance will fetch the remaining archived redo log files.

12.9.2 Standby Database Does Not Contain Data Files, Used as a FAL Server

Use the same procedure described in Section 12.4, with the exception that the RMAN commands that back up database files cannot be run against the FAL server. The FAL server can be used as a backup source for all archived redo log files, thus off-loading backups of archived redo log files to the FAL server.

12.9.3 Standby Database File Names Are Different From Primary Database

If the database filenames are not the same on the primary and standby databases that were never resynchronized, the RESTORE and RECOVER commands you use will be slightly different. To obtain the actual data file names on the standby database, query the V$DATAFILE view and specify the SET NEWNAME option for all the data files in the database:

RUN 
{
SET NEWNAME FOR DATAFILE 1 TO 'existing file location for file#1 from V$DATAFILE';
SET NEWNAME FOR DATAFILE 2 TO 'existing file location for file#2 from V$DATAFILE';
…
…
 SET NEWNAME FOR DATAFILE n TO 'existing file location for file#n from V$DATAFILE';
 RESTORE {DATAFILE <n,m,…> | TABLESPACE tbs_name_1, 2, …| DATABASE;
SWITCH DATAFILE ALL; 
RECOVER DATABASE {NOREDO};
}

Similarly, the RMAN DUPLICATE command should also use the SET NEWNAME option to specify new filenames during standby database creation. Or you could set the LOG_FILE_NAME_CONVERT and DB_FILE_NAME_CONVERT parameters.