7
Making Backups wi th RMAN: Advanced Topics

When you create multiple backup sets and allocate multiple channels , RMAN automatically writes multiple backup sets in parallel. The allocated server sessions share the work of backing up the specifie d datafiles, control files, and archived redo logs. Note that you cannot stripe a single backup set across multiple channels.

RMAN automatically assigns a backup set to a device. You can use the CHANNEL paramete r so that RMAN writes all backup sets for a backupSpec to a specific channel.

For example, this example parallelizes the backup operation by specifying which channels RMAN should back up to disk and whi ch to sbt:

RMAN> RUN
{
  ALLOCA
TE CHANNEL ch1 DEVICE TYPE DISK FORMAT = '/backup/df/%U';
  ALLOCATE CHANNEL ch2 DEVICE TYPE DISK FORMAT = '/ba
ckup/cf/%U';
  ALLOCATE CHANNEL ch3 DEVICE TYPE sbt;
  BACKUP AS BACKUPSET # all output f
iles are in backup sets
    # channel ch1 backs up datafiles to /backup/df directory

DATAFILE 1,2,3,4 
    CHANNEL ch1
    # channel ch2 backs up control file copy to /backup
/cf directory
    CONTROLFILECOPY '/tmp/control01.ctl'
    CHANNEL ch2;
   BACKUP AS BACKUPSET
    # channel ch3 backs up archived redo logs to tape
    ARC
HIVELOG FROM TIME 'SYSDATE-14'
    CHANNEL ch3;
}

You cannot back up to DISK and sbt at the same time using automatic channels: you must manually allocate them.

Backing Up in NOARCHIVELOG Mode: Example

Thi s script puts the database into the correct mode for a consistent, whole database backup and then backs up the database. The script p erforms a shutdown, startup, shutdown, and then startup again before creating multiple copies of the backup:

# Shut down database cleanly with immediate option. This type of shutdown lets  
# current ca
lls to the database complete, but prevents further logons or calls. 
# If the database is not up, you receive a
 message saying so but RMAN will not 
# treat this situation as an error.
SHUTDOWN IMMEDI
ATE; 
  
# Start up the database in case it suffered instance failure or was 
# closed with SHUTDOWN ABORT before starting this script. 
# The script performs instance recovery i
f 
# needed. Oracle uses the default init.ora file. Alternatively, use this form:
# START
UP FORCE DBA pfile=filename. 
# Use the DBA option because you are going to shut down again
# and do not want to let users in during the short interval. Use the FORCE 
# option because it can
not hurt and might help in certain situations. 
STARTUP FORCE DBA; 
SHUTDOWN IMMEDIATE; 
  
# The database is cleanly closed and ready for a consistent backup. RMAN
# requires that the database be started and mounted to perform a backup.
RMAN> STARTUP MOUNT;

# this example uses automatic channels to make the backup
BACKUP 
  COPIES 2 
  INCREMENTAL LEVEL 0 
  MAXSETSIZE 10M 
<
/a>  DATABASE;

# Now that the backup is complete, open the database. 

ALTER DATABASE OPEN; 

You can skip tablespaces, but any skipp ed tablespace that has not been offline or read-only since its last backup will be lost if the database has to be restored from a bac kup. When backing up to disk, make sure that the destination file system has enough free space.

Cataloging User-Managed Datafile Copies: Examp le

You can use operating system utilities to make copies of datafile s and then catalog them in the recovery catalog.

If the database is open and the datafile i s online, then issue ALTER TABLESPACE ... BEGIN BACKUP to put the da tafile in backup mode. (If you will be copying all datafiles, you can use ALTER DATABASE BEGIN BACKUP to put all tablesp aces in backup mode, instead of issuing individual commands for each tablespace.) Then copy the files using native operating system c ommands. Once the copy is completed, use ALTER TABLESPACE... END BACKUP to take the datafile out of backup mode (or use ALTER DATABASE END BACKUP to take all tablespaces in backup mode out of backup mode). Finally, catalog the resulting datafile copy using the RMAN CATALOG DATAFILECOPY command.

For examp le, if you created a copy of your datafile /oracle/oradata/users01.dbf into file /tmp/users01.dbf, this command will add /tmp/users01 .dbf to the catalog:

CATALOG DATAFILECOPY '/tmp/users01.dbf';

If you try to catalog a datafile copy from a database other than the connected target database , then RMAN issues an error such as the following:

RMAN-00571: ============================
===============================
RMAN-00569: =============== ERROR MESSAGE STACK FOLLOWS ===============
RMAN-00571: ===========================================================
RMAN-03009: failure of c
atalog command on default channel at 08/29/2001 14:44:34
ORA-19563: datafile copy header validation failed for
file /tmp/tools01.dbf

Keeping a Long-Term Backup: Example

If you configure a retention policy, then you may want to exclude specified backups from this policy. For example, you may want to archive a consiste nt backup of the database once a year to serve as a historical record. This long-term backups does not function as a backup that you may perform recovery on, but an archived snapshot of data at a particular time.

To exempt a backup from the retention policy, specify the KEEP option on the BACKUP command. You can also specify LOGS or NOLOGS to indicate whether RMAN should save archived logs for possible recovery of this backup. If you specify NOLOGS, then the backup must be consistent.

This example keeps the bac kup of the database indefinitely and does not save archived logs needed to recover it:

RMAN
> SHUTDOWN IMMEDIATE;
RMAN> STARTUP MOUNT;  # put database in consistent state
RMAN
> BACKUP DATABASE KEEP FOREVER NOLOGS 
              TAG 'db_archive_1'; # make long-term consistent backup

# mark backup as unavailable in the repository so that RMAN does not attempt to
# restore it unless explicitly specified on the RESTORE command
RMAN> CHANGE BACKUP TAG 'db_ar
chive_1' UNAVAILABLE;
RMAN> SQL 'ALTER DATABASE OPEN';

Optimizing Backups: Examples

Run the CONFIGURE BACKUP OPTIMIZATION command to enable bac kup optimization. When specific conditions are met (described in "Backup Optimization Algorithm"), RMAN skips backups of files that are identical to files that are already backed up.

Assume that you configure optimization and a retention policy as follows:

CONFIGURE DEFAULT DEVICE TYPE TO sbt;
CONFIGURE BACKUP OPTIMIZATION ON;
CONFI
GURE RETENTION POLICY TO RECOVERY WINDOW OF 4 DAYS;

Optimizing a Database Backup: Example

Then, you run this command every night to back up the database to tape:

BAC
KUP DATABASE;

Because backup optimization is configured, RMAN ski ps backups of offline and read-only datafiles only if the most recent backups were made on or after the earliest point in the recover y window. RMAN does not skip backups when the most recent backups are older than the window. For example, optimization ensures you do not end up with a new backup of the read-only datafile ?/oradata/trgt/history01.dbf every night, so long as one backup set containing this file exists within the recovery window.

For example, if the most recent backup of the datafiles was on Sunday, and the point of recoverability (that is, the earliest date in the recovery window) is on Sat urday, then RMAN skips the datafiles when you run the Wednesday backup. On Friday, the point of recoverability is now Monday, so the Sunday backup is now outside the window. Hence, the Friday backup does not skip the datafiles.

Optimizing a Daily Archived Log Backup to a Sin gle Tape: Example

Assume that you want to back up all the archived l ogs every night. However, you do not want to have multiple copies of each log sequence number. So, you configure backup optimization to ON, then run this command in a script every night at 1 a.m.:

BACKUP DEVICE
TYPE sbt ARCHIVELOG ALL;

RMAN skips all logs except those produce d in the last 24 hours. In this way, you keep only one copy of each archived log on tape.

Optimizing a Daily Archived Log Backup to Multiple T apes: Example

In this example, you back up logs that are not already on tape to one tape pool, then back up the same logs to a second tape pool. Finally, you delete old logs.

For the first step, perform the one-time configuration:

# configure backup
optimization
CONFIGURE BACKUP OPTIMIZATION ON;
CONFIGURE DEFAULT DEVICE TYPE TO sbt;

Then, run the following script at the same time every night to back u p the logs generated during the previous day to two separate tape pools:

# The following co
mmand backs up just the logs that are not on tape. The 
# first copies are saved to the tapes from the pool "ar
chivelog_pool_1"
RUN
{
  ALLOCATE CHANNEL c1 DEVICE TYPE sbt
    PARMS='NSR_DATA_VOLUME_POOL=ARCHIVELOG_POOL_1';
  BACKUP ARCHIVELOG ALL;

}
# Make one more copy of the archived logs and save them to tapes from a 
# differen
t pool 
RUN
{
  ALLOCATE CHANNEL c2 DEVICE TYPE sbt 
    PARMS='NSR_DATA_VOLUME_POOL=ARCHIVELOG_POOL_2';
  BACKUP ARCHIVELOG
    FROM
TIME 'SYSDATE-1'
    UNTIL TIME 'SYSDATE'; # specify UNTIL so RMAN does not archive current log
}
# Delete old logs - for example, delete logs created within the last week. 
DELE
TE ARCHIVELOG ALL COMPLETED AFTER 'SYSDATE-7';

Creating a Weekly Secondary Backup of Archived Logs: Example

Assume a more sophisticated scenario in which your goal is to back up the archived logs to tape eve ry day. However, you are worried about tape failure, so you want to ensure that you have more than copy of each log sequence number o n an separate tape before you perform your weekly deletion of logs from disk.

First, perfor m a one-time configuration:

# configure backup optimization
CONFIGURE
 BACKUP OPTIMIZATION ON;
CONFIGURE DEVICE TYPE sbt PARALLELISM 1;
CONFIGURE default DEVIC
E TYPE TO sbt;
# configure a default channel that sends backups to tape pool "first_copy"
CONFIGURE CHANNEL DEVICE TYPE sbt PARMS='ENV=(NSR_DATA_VOLUME_POOL=first_copy);

Because you have optimization enabled, you can run the following command every evening to back up all archived logs t o the "first_copy" pool that have not already been backed up:

BACKUP ARCHIVELOG ALL TAG fir
st_copy;

Every Friday evening you create an additional backup of all archived logs in a different tape pool. Also, at the end of the backup, you want to delete all archived logs that already have at least two copies on tape. So you run the following script:

RUN
{
  # manually allocate a channel, in order to specify that the backup run by this
  # channel
 should go to both pools "first_copy" and "second_copy"
  ALLOCATE CHANNEL c1 DEVICE TYPE sbt
      PARMS='ENV=(NSR_DATA_VOLUME_POOL=second_copy)';
  ALLOCATE CHANNEL c2 DEVICE TYPE sbt
      PARMS='ENV=(NSR_DATA_VOLUME_POOL=first_copy)';
  BACKUP CHANNEL C1 ARCHIVELOG UNTIL TIME 'SYSDATE
' 
    NOT BACKED UP 2 TIMES # back up only logs without 2 backups on tape
    TAG SECOND
_COPY; 
  BACKUP CHANNEL C2 ARCHIVELOG UNTIL TIME 'SYSDATE' 
    NOT BACKED UP 2 TIMES #
back up only logs without 2 backups on tape
    TAG FIRST_COPY;
}

# now delete from disk all logs that have been backed up to tape at least twice
DELETE AR
CHIVELOG ALL
  BACKED UP 2 TIMES TO DEVICE TYPE sbt;

The Friday script creates a second copy of all archived logs in the "second_copy" tape pool. After the backup, you can send the tape from the pool "second_copy" to secure storage. You should use this tape backup only if the primary tape from pool "first_co py" is damaged. Because the secondary tape is in a secure place, you do not want RMAN to use it for recovery, so you can mark the bac kup as unavailable:

CHANGE BACKUP OF ARCHIVELOG TAG SECOND_COPY UNAVAILABLE;

Handling Errors Duri ng Backups: Example

By default a checksum is calculated for every bl ock read from a datafile and stored in the backup or image copy. If you use the NOCHECKSUM option, then checksums are no t calculated. If the block already contains a checksum, however, then the checksum is validated and stored in the backup. If the vali dation fails, then the block is marked corrupt in the backup.

The SET MA XCORRUPT FOR DATAFILE command sets how many corrupt blocks in a datafile that BACKUP wi ll tolerate. If a datafile has more corrupt blocks than specified by the MAXCORRUPT parameter, the command terminates. I f you specify the CHECK LOGICAL option, RMAN checks for logical and physical corruption.

By default, the BACKUP command terminates when it cannot access a datafile. You can specify para meters to prevent termination, as listed in the following table.

The following example uses an automatic cha nnel to back up the database, and sets the corruption level for the datafile in the SYSTEM tablespace so that up to 10 e rrors will be accepted:

RMAN> RUN
{
  SET MA
XCORRUPT FOR DATAFILE 1 TO 10;
  BACKUP DATABASE
    SKIP INACCESSIBLE

    SKIP READONLY
    SKIP OFFLINE; 
}