Rman, duplicate from incrementally updated backup

In the last post i addressed how to create and maintain an incrementally updated rman backup, and how to switch to the backup, in case some corruption happens to the production database.

In this post, instead of doing the switch, we will use the incrementally updated backup to create a cloned copy of the original database. The practical use of this method will be that, if you are creating incrementally updated backups every night, you can use the backup as the source to clone the database (ie you dont have to run a whole backup of your database again).

rman has a command named duplicate that allows you to clone an existing database.

In this case we are cloning a database named rk01, to another database named rk02 on the same host, using a incrementally updated backup, with the rman duplicate command.

First follow the steps to create the incrementally updated backup from my previous post.

Enable Archivelog And Block Change Tracking
Run a level 0 full image copy backup of the database
Perform updates in the database
Perform an incremental level 1 backup
Apply the incremental level 1 backup to the level 0 image copy

At this point after the incrementally updated backup is created, in the previous post, i was doing a switch. Instead of that we will do the rman duplicate.

Create a spfile for the rk02 database (I usually make a pfile from spfile from rk01, edit the pfile change all the occurances of rk01 to rk02 and create the spfile for rk02 from this pfile. Also create the necessary directories for bdump,adump etc.)

Create a password file for the rk02 database (orapwd file=$ORACLE_HOME/dbs/orapwrk02 entries=30)

Startup the database rk02 in nomount mode

export ORACLE_SID=rk02

sqlplus / as sysdba

startup nomount

Switch the current logfile in the rk01 database

This is an important step. Miss this and your rman duplicate will likely fail. The reason is that the scn until which the incremental backup is taken (And was applied to the original backup copy), is still in the online redo log. When you do a duplicate rman looks for a backup that has a ckp scn that is < the first_change# in your current online redo log. Once you do a switch, and then run the duplicate, rman will be able to find the incrementally updated backup as the source to clone from.

export ORACLE_SID=rk01

sqlplus / as sysdba

alter system archive log current;

Perform the rman duplicate.

Then issue the rman duplicate command (If you are cloning to a physically separate host, you will have to create appropriate listener.ora entries, tnsnames.ora entries and make the location of the backup copy available to your target host).

export ORACLE_SID=rk02

run the following script.

#!/bin/bash
rman target="sys/manager@rk01" auxiliary / log=/u01/oraback/clonefromiub.log << EOF
RUN
{
  SET NEWNAME FOR DATAFILE 1 TO '/u01/oradata/rk02/system01.dbf';
  SET NEWNAME FOR DATAFILE 2 TO '/u01/oradata/rk02/sysaux01.dbf';
  SET NEWNAME FOR DATAFILE 3 TO '/u01/oradata/rk02/undotbs01.dbf';
  SET NEWNAME FOR DATAFILE 4 TO '/u01/oradata/rk02/users01.dbf';
  SET NEWNAME FOR DATAFILE 5 TO '/u01/oradata/rk02/example01.dbf';
  SET NEWNAME FOR TEMPFILE 1 TO '/u01/oradata/rk02/temp01.dbf';
  DUPLICATE  TARGET DATABASE  TO rk02
    LOGFILE
      GROUP 1 ('/u01/oradata/rk02/redo01a.log',
               '/u01/oradata/rk02/redo01b.log') SIZE 4M REUSE,
      GROUP 2 ('/u01/oradata/rk02/redo02a.log',
               '/u01/oradata/rk02/redo02b.log') SIZE 4M REUSE;
}
EXIT;
EOF