Oracle DBA: Backup and Recovery Scripts
Having the right backup and recovery procedures is the lifeblood of any database. Companies live on data, and, if that data is not available, the whole company collapses. As a result, it is the responsibility of the database administrator to protect the database from system faults, crashes, and natural calamities resulting from a variety of circumstances.
The choice of a backup and recovery mechanism depends mainly on the following factors:
Database mode (ARCHIVELOG, NOARCHIVELOG)
Size of the database
Backup and recovery time
Type of data (OLTP, DSS, Data Warehouse).
The types of backup are
Offline backup (Cold or closed database backup)
Online backup (Hot or open database backup)
Logical exports create an export file that contains a list of SQL statements to recreate the database. Export is performed when the database is open and does not affect users work. Offline backups can only be performed when the database is shut down cleanly, and the database will be unavailable to users while the offline backup is being performed. Online backups are performed when the database is open, and it does not affect users work. The database needs to run in ARCHIVELOG mode to perform online backups.
The database can run in either ARCHIVELOG mode or NOARCHIVELOG mode. In ARCHIVELOG mode, the archiver (ARCH) process archives the redo log files to the archive destination directory. These archive files can be used to recover the database in the case of a failure. In NOARCHIVELOG mode, the redo log files are not archived.
When the database is running in ARCHIVELOG mode, the choice can be one or more of the following:
When the database is running in NOARCHIVELOG mode, the choice of backup is as follows:
Offline or cold backups are performed when the database is completely shutdown. The disadvantage of an offline backup is that it cannot be done if the database needs to be run 24/7. Additionally, you can only recover the database up to the point when the last backup was made unless the database is running in ARCHIVELOG mode.
The general steps involved in performing a cold backup are shown in Figure 3.1. These general steps are used in writing cold backup scripts for Unix and Windows NT.
Figure 3.1 Steps for cold backup.
The steps in Figure 3.1 are explained as follows.
Step 1Generating File List
An offline backup consists of physically copying the following files:
Init.ora and config.ora files
Backing up online redo log files is not advised in all cases, except when performing cold backup with the database running in NOARCHIVELOG mode. If you make a cold backup in ARCHIVELOG mode do not backup redo log files. There is a chance that you may accidentally overwrite your real online redo logs, preventing you from doing a complete recovery.
If your database is running in ARCHIVELOG mode, when you perform cold backup you should also backup archive logs that exist.
Before performing a cold backup, you need to know the location of the files that need to be backed up. Because the database structure changes day to day as more files get added or moved between directories, it is always better to query the database to get the physical structure of database before making a cold backup.
To get the structure of the database, query the following dynamic data dictionary tables:
- V$datafile Lists all the data files used in the database
SQL>select name from v$datafile;
- Backup the control file and perform a trace of the control file using
SQL>alter database backup controlfile to '/u10/backup/control.ctl';
SQL>alter database backup controlfile to trace;
Init.ora and config.ora Located under $ORACLE_HOME/dbs directory
Step 2Shut down the database
You can shut down a database with the following commands:
$su oracle $sqlplus "/ as sysdba" SQL>shutdown
Step 3Perform a backup
In the first step, you generated a list of files to be backed up. To back up the files, you can use the Unix copy command (cp) to copy it to a backup location, as shown in the following code. You have to copy all files that you generated in Step 1.
$cp /u01/oracle/users01.dbf /u10/backup
You can perform the backup of the Init.ora and config.ora files as follows:
$cp $ORACLE_HOME/dbs/init.ora /u10/backup $cp $ORACLE_HOME/dbs/config.ora /u10/backup
Step 4Start the database
After the backup is complete, you can start the database as follows:
$su oracle $sqlplus "/ as sysdba" SQL> startup
An online backup or hot backup is also referred to as ARCHIVE LOG backup. An online backup can only be done when the database is running in ARCHIVELOG mode and the database is open. When the database is running in ARCHIVELOG mode, the archiver (ARCH) background process will make a copy of the online redo log file to archive backup location.
An online backup consists of backing up the following files. But, because the database is open while performing a backup, you have to follow the procedure shown in Figure 3.2 to backup the files:
Data files of each tablespace
Archived redo log files
Init.ora and config.ora files
Figure 3.2 Steps for hot backup.
The general steps involved in performing hot backup are shown in Figure 3.2. These general steps are used in writing hot backup scripts for Unix and Windows NT.
The steps in Figure 3.2 are explained as follows.
Step 1Put the tablespace in the Backup mode and copy the data files.
Assume that your database has two tablespaces, USERS and TOOLS. To back up the files for these two tablespaces, first put the tablespace in backup mode by using the ALTER statement as follows:
SQL>alter tablespace USERS begin backup;
After the tablespace is in Backup mode, you can use the SELECT statement to list the data files for the USERS tablespace, and the copy (cp) command to copy the files to the backup location. Assume that the USERS tablespace has two data filesusers01.dbf and users02.dbf.
SQL>select file_name from dba_data_files where tablespace_name='USERS'; $cp /u01/oracle/users01.dbf /u10/backup $cp /u01/oracle/users01.dbf /u10/backup
The following command ends the backup process and puts the tablespace back in normal mode.
SQL>alter tablespace USERS end backup;
You have to repeat this process for all tablespaces. You can get the list of tablespaces by using the following SQL statement:SQL>select tablespace_name from dba_tablespaces;
Step 2Back up the control and Init.ora files.
To backup the control file,
SQL>alter database backup controlfile to '/u10/backup/control.ctl';
You can copy the Init.ora file to a backup location using
$cp $ORACLE_HOME/dbs/initorcl.ora /u10/backup
Step 3Stop archiving.
Archiving is a continuous process and, without stopping archiver, you might unintentionally copy the file that the archiver is currently writing. To avoid this, first stop the archiver and then copy the archive files to backup location. You can stop the archiver as follows:
SQL>alter system switch logfile; SQL>alter system archive log stop;
The first command switches redo log file and the second command stops the archiver process.
Step 4Back up the archive files.
To avoid backing up the archive file that is currently being written, we find the least sequence number that is to be archived from the V$LOG view, and then backup all the archive files before that sequence number. The archive file location is defined by the LOG_ARCHIVE_DEST_n parameter in the Init.ora file.
select min(sequence#) from v$log where archived='NO';
Step 5Restart the archive process.
The following command restarts the archiver process:
SQL>alter system archive log start;
Now you have completed the hot backup of database.
An online backup of a database will keep the database open and functional for 24/7 operations. It is advised to schedule online backups when there is the least user activity on the database, because backing up the database is very I/O intensive and users can see slow response during the backup period. Additionally, if the user activity is very high, the archive destination might fill up very fast.
Database Crashes During Hot Backup
There can be many reasons for the database to crash during a hot backupa power outage or rebooting of the server, for example. If these were to happen during a hot backup, chances are that tablespace would be left in backup mode. In that case you must manually recover the files involved, and the recovery operation would end the backup of tablespace. It's important to check the status of the files as soon as you restart the instance and end the backup for the tablespace if it's in backup mode.
select a.name,b.status from v$datafile a, v$backup b where a.file#=b.file# and b.status='ACTIVE';
select a.tablespace_name,a.file_name,b.status from dba_data_files a, v$backup b where a.file_id=b.file# and b.status='ACTIVE';
This statement lists files with ACTIVE status. If the file is in ACTIVE state, the corresponding tablespace is in backup mode. The second statement gives the tablespace name also, but this can't be used unless the database is open. You need to end the backup mode of the tablespace with the following command:
alter tablespace tablespace_name end backup;
Export is the single most versatile utility available to perform a backup of the database, de-fragment the database, and port the database or individual objects from one operating system to another operating system.
Export backup detects block corruption
Though you perform other types of backup regularly, it is good to perform full export of database at regular intervals, because export detects any data or block corruptions in the database. By using export file, it is also possible to recover individual objects, whereas other backup methods do not support individual object recovery.
Export can be used to export the database at different levels of functionality:
Full export (full database export) (FULL=Y)
User-level export (exports objects of specified users) (OWNER=userlist)
Table-level export (exports specified tables and partitions) (TABLES=tablelist)
Transportable tablespaces (TABLESPACES=tools, TRANSPORT_TABLESPACE=y)
There are two methods of Export:
Conventional Path (default)Uses SQL layer to create the export file. The fact is that the SQL layer introduces CPU overhead due to character set, converting numbers, dates and so on. This is time consuming.
Direct path (DIRECT=YES)Skips the SQL layer and reads directly from database buffers or private buffers. Therefore it is much faster than conventional path.
We will discuss scripts to perform the full, user-level, and table-level export of database. The scripts also show you how to compress and split the export file while performing the export. This is especially useful if the underlying operating system has a limitation of 2GB maximum file limit.
This chapter requires understanding of basic Unix shell and DOS batch programming techniques that are described in Chapter 2 "Building Blocks." That chapter explained some of the common routines that will be used across most of the scripts presented here.
This book could have provided much more simple scripts. But, considering standardization across all scripts and the reusability of individual sections for your own writing of scripts, I am focusing on providing a comprehensive script, rather than a temporary fix. After you understand one script, it is easy to follow the flow for the rest of the scripts.