Oracle Backup and Recovery

Backup and recovery is one of the most important aspects of database administration. If a database crashed and there was no way to recover it, the devastating results to a business could include lost data, lost revenue and customer dissatisfaction. Whether companies operate a single database or multiple databases storing hundreds of gigabytes or even terabytes of data, they share one common factor - the need to back up important data and protect themselves from disaster by developing a backup and recovery plan.

BASICS OF BACKUP AND RECOVERY
A backup is a representative copy of data. This copy can include important parts of a database such as the control file, redo logs, and datafiles. A backup protects data from application error and acts as a safeguard against unexpected data loss, by providing a way to restore original data. Backups are divided into physical backups and logical backups. Physical backups are copies of physical database files. The phrase "backup and recovery" usually refers to the transfer of copied files from one location to another, along with the various operations performed on these files.

In contrast, logical backups contain data that is exported using SQL commands and stored in a binary file. Oracle records both committed and uncommitted changes in redo log buffers. Logical backups are used to supplement physical backups. Restoring a physical backup means reconstructing it and making it available to the Oracle server. To recover a restored backup, data is updated using redo records from the transaction log. The transaction log records changes made to the database after the backup was taken.

Oracle performs crash recovery and instance recovery automatically after an instance failure. In the case of media failure, a database administrator (DBA) must initiate a recovery operation. Recovering a backup involves two distinct operations: rolling the backup forward to a more recent time by applying redo data, and rolling back all changes made in uncommitted transactions to their original state. In general, recovery refers to the various operations involved in restoring, rolling forward, and rolling back a backup. Backup and recovery refers to the various strategies and operations involved in protecting the database against data loss and reconstructing the database should a loss occur.

BACKUP AND RECOVERY OPERATIONS
A backup is a snapshot of a datafile, tablespace, or database at a certain time. If periodic backups of the database have been made and data is lost, users can apply the stored redo information to their latest backup to make the database current again. Oracle enables users to restore an older backup and apply only some redo data, thereby recovering the database to an earlier point in time. This type of recovery is called incomplete media recovery. If the backup was consistent, then users are not required to apply any redo data, at all.

A simple example of media recovery illustrates the concept. Suppose a user makes a backup of the database at noon. Starting at noon, one change to the database is made every minute. At 1 p.m. one of the disk drives fails, causing the loss of all data on that disk. Fortunately, Oracle records all changes in the redo log. The user can then restore the noon backup onto a functional disk drive and use redo data to recover the database to 1 p.m., reconstructing the lost changes.

ELEMENTS OF A BACKUP AND RECOVERY STRATEGY
Although backup and recovery operations can be intricate and vary from one business to another, the basic principles follow these four simple steps:

1. Multiplex the online redo logs
2. Run the database in ARCHIVELOG mode and archive redo logs to multiple locations
3. Maintain multiple concurrent backups of the control file
4. Take frequent backups of physical datafiles and store them in a safe place, making multiple copies if possible

As long as users have backups of the database and archive redo logs in safe storage, the original database can be recreated.

KEY DATA STRUCTURES FOR BACKUP AND RECOVERY
Before users begin to think seriously about backup and recovery strategy, the physical data structures relevant for backup and recovery operations must be identified. This section discusses the following physical data structures:

• Datafiles
• Control Files
• Online Redo Log Files
• Archived Redo Log Files
• Automatic Managed Undo

DATAFILES
Every Oracle database has one or more physical datafiles that belong to logical structures called tablespaces. The datafile is divided into smaller units called data blocks. The data of logical database structures, such as tables and indexes, is physically located in the blocks of the datafiles allocated for a database. Datafiles hold the following characteristics:

• User-defined characteristics allow datafiles to automatically extend when the database runs out of space.
• One or more physical datafiles form a logical database storage unit called a tablespace.

The first block of every datafile is the header. The header includes important information such as file size, block size, tablespace, and creation timestamp. Whenever the database is opened, Oracle checks to see that the datafile header information matches the information stored in the control file. If it does not, then recovery is necessary. Oracle reads the data in a datafile during normal operation and stores it in the buffer cache. For example, assume that a user wants to access some data in a table. If the requested information is not already in the buffer cache, Oracle reads it from the appropriate datafiles and stores it in memory.

CONTROL FILES
Every Oracle database has a control file containing the operating system filenames of all other files that constitute the database. This important file also contains consistency information that is used during recovery, such as the:

• Database name
• Timestamp of database creation
• Names of the database's datafiles and online and archived redo log files
• Checkpoint, a record indicating the point in the redo log where all database changes prior to this point have been saved in the datafiles
• Recovery Manager(RMAN) backup meta-data

Users can multiplex the control file, allowing Oracle to write multiple copies of the control file to protect it against disaster. If the operating system supports disk mirroring, the control file can also be mirrored, allowing the O/S to write a copy of the control file to multiple disks. Every time a user mounts an Oracle database, its control file is used to identify the datafiles and online redo log files that must be opened for database operation. If the physical makeup of the database changes, such as a new datafile or redo log file is created, Oracle then modifies the database's control file to reflect the change. The control file should be backed up whenever the structure of the database changes. Structural changes can include adding, dropping, or altering datafiles or tablespaces and adding or dropping online redo logs.

ONLINE REDO LOG FILES
Redo logs are absolutely crucial for recovery. For example, imagine that a power outage prevents Oracle from permanently writing modified data to the datafiles. In this situation, an old version of the data in the datafiles can be combined with the recent changes recorded in the online redo log to reconstruct what was lost. Every Oracle database contains a set of two or more online redo log files. Oracle assigns every redo log file a log sequence number to uniquely identify it. The set of redo log files for a database is collectively known as the database's redo log.

Oracle uses the redo log to record all changes made to the database. Oracle records every change in a redo record, an entry in the redo buffer describing what has changed. For example, assume a user updates a column value in a payroll table from 5 to 7. Oracle records the old value in undo and the new value in a redo record. Since the redo log stores every change to the database, the redo record for this transaction actually contains three parts:

• The change to the transaction table of the undo
• The change to the undo data block
• The change to the payroll table data block

If the user then commits the update to the payroll table - to make permanent the changes executed by SQL statements - Oracle generates another redo record. In this way, the system maintains a careful watch over everything that occurs in the database.

Circular Use of Redo Log Files
Log Writer (LGWR) writes redo log entries to disk. Redo log data is generated in the redo log buffer of the system global area. As transactions commit and the log buffer fills, LGWR writes redo log entries into an online redo log file. LGWR writes to online redo log files in a circular fashion: when it fills the current online redo log file, called the active file, LGWR writes to the next available inactive redo log file. LGWR cycles through the online redo log files in the database, writing over old redo data. Filled redo log files are available for reuse depending on whether archiving is enabled:

• If archiving is disabled, a filled online redo log is available once the changes recorded in the log have been saved to the datafiles.
• If archiving is enabled, a filled online redo log is available once the changes have been saved to the datafiles and the file has been archived.

ARCHIVED REDO LOG FILES
Archived log files are redo logs that Oracle has filled with redo entries, rendered inactive, and copied to one or more log archive destinations. Oracle can be run in either of two modes:

• ARCHIVELOG - Oracle archives the filled online redo log files before reusing them in the cycle.
• NOARCHIVELOG - Oracle does not archive the filled online redo log files before reusing them in the cycle.

Running the database in ARCHIVELOG mode has the following benefits:

• The database can be completely recovered from both instance and media failure.
• The user can perform online backups, i.e., back up tablespaces while the database is open and available for use.
• Archived redo logs can be transmitted and applied to the standby database
• Oracle supports multiplexed archive logs to avoid any possible single point of failure on the archive logs.
• The user has more recovery options, such as the ability to perform tablespace-point-in-time recovery (TSPITR)

Running the database in NOARCHIVELOG mode has the following consequences:

• The user can only back up the database while it is completely closed after a clean shutdown.
• Typically, the only media recovery option is to restore the whole database, which causes the loss of all transactions issued since the last backup.

Automatic Managed Undo
Every Oracle database must have a method of maintaining information that is used to roll back, or undo, changes to the database. Such information consists of records of the actions of transactions, primarily before they are committed. Oracle refers to these records collectively as undo. Historically, Oracle has used rollback segments to store undo. Space management for these rollback segments has proven to be quite complex. Oracle now offers another method of storing undo that eliminates the complexities of managing rollback segment space, and allows DBAs to exert control over how long undo is retained before being overwritten. This method uses an undo tablespace. Undo records are used to:

• Roll back transactions when a ROLLBACK statement is issued
• Recover the database
• Provide read consistency

When a rollback statement is issued, undo records are used to undo changes that were made to the database by the uncommitted transaction. During database recovery, undo records are used to undo any uncommitted changes applied from the redo log to the datafiles. Undo records provide read consistency by maintaining the before image of the data for users who are accessing the data at the same time that another user is changing it.

Understanding Basic Backup
A backup strategy provides a safeguard against data loss. Answering the following questions can help database administrators develop a strong backup strategy:

• What types of failures can occur?
• What information should be backed up?
• Which backup method should be used?
• Should backups be made online or offline?
• How often should backups be made?
• How can dangerous backup techniques be avoided?

WHAT TYPES OF FAILURES CAN OCCUR?
Data loss can occur for various reasons. Here are some of the most common types of failures that can lead to data loss.

A statement failure is a logical failure in the handling statement in an Oracle program. For example, a user issues a statement that is not a valid SQL construction. When statement failure occurs, Oracle automatically undoes any effects of the statement and returns control to the user.

A process failure is a failure in a user process accessing Oracle, i.e., an abnormal disconnection or process termination. The failed user process cannot continue work, although Oracle and other user processes can. If the user process fails while modifying the database, Oracle background processes undo the effects of uncommitted transactions.

An instance failure is a problem that prevents an Oracle instance, i.e., the SGA and background processes, from continuing to function. Instance failure can result from a hardware problem such as a power outage, or a software problem such as an operating system crash. When an instance fails, Oracle does not write the data in the buffers of the SGA to the datafiles.

A user or application error is a user mistake that results in the loss of data. For example, a user can accidentally delete data from a payroll table. Such user errors can require the database or object to be recovered to a point in time before the error occurred. To allow recovery from user error and accommodate other unique recovery requirements, Oracle provides Flashback Technology. For more information on Flashback Technology.

A media failure is a physical problem that arises when Oracle tries to write or read a file that is required to operate the database. A common example is a disk head crash that causes the loss of all data on a disk drive. Disk failure can affect a variety of files, including datafiles, redo log files, and control files. Because the database instance cannot continue to function properly, it cannot write the data in the database buffers of the SGA to the datafiles.

WHAT INFORMATION SHOULD BE BACKED UP?
A database contains a wide variety of types of data. When developing backup strategy, DBAs must decide what information they want to copy. The basic backup types include:

• Online Database Backup
• Offline Database Backup
• Whole Database
• Tablespace
• Datafile
• Control File
• Archived Redo Log
• Configuration Files

In deciding what to back up, the basic principle is to prioritize data depending on its importance and the degree to which it changes. Archive logs do not change, for example, but they are crucial for recovering the database, so multiple copies should be maintained, if possible. Expense account tables, however, are constantly updated by users. Therefore, this tablespace should be backed up frequently to prevent having to apply as much redo data during recovery.

Backups can be combined in a variety of ways. For example, a DBA can decide to take weekly whole database backups, to ensure a relatively current copy of original database information, but take daily backups of the most accessed tablespaces. The DBA can also multiplex the all important control file and archived redo log as an additional safeguard.

Online Database Backup
An online backup or also known as an open backup, is a backup in which all read-write datafiles and control files have not been checkpointed with respect to the same SCN. For example, one read-write datafile header may contain an SCN of 100 while other read-write datafile headers contain an SCN of 95 or 90. Oracle cannot open the database until all of these header SCNs are consistent, that is, until all changes recorded in the online redo logs have been saved to the datafiles on disk. If the database must be up and running 24 hours a day, 7 days a week, then you have no choice but to perform online backups of a whole database which is in ARCHIVELOG mode.

Offline Database Backup
In this backup, all datafiles and control files are consistent to the same point in time - consistent with respect to the same SCN, for example. The only tablespaces in a consistent backup that are allowed to have older SCNs are read-only and offline-normal tablespaces, which are consistent with the other datafiles in the backup. This type of backup allows the user to open the set of files created by the backup without applying redo logs, since the data is already consistent. The only way to perform this type of backup is to shut down the database cleanly and make the backup while the database is closed. A consistent whole database backup is the only valid backup option for databases running in NOARCHIVELOG mode.

Whole Database Backup
The most common type of backup, a whole database backup contains the control file along with all database files that belong to a database. If operating in ARCHIVELOG mode, the DBA also has the option of backing up different parts of the database over a period of time, thereby constructing a whole database backup piece by piece.

Tablespace Backups
A tablespace backup is a subset of the database. Tablespace backups are only valid if the database is operating in ARCHIVELOG mode. The only time a tablespace backup is valid for a database running in NOARCHIVELOG mode is when that tablespace is read-only or offline-normal.

Datafile Backups
A datafile backup is a backup of a single datafile. Datafile backups, which are not as common as tablespace backups and are only valid if the database is run in ARCHIVELOG mode. The only time a datafile backup is valid for a database running in NOARCHIVELOG mode is if that datafile is the only file in a tablespace. For example, the backup is a tablespace backup, but the tablespace only contains one file and is read-only or offline-normal.

Control File Backups
A control file backup is a backup of a database's control file. If a database is open, the user can create a valid backup by issuing the following SQL statement: ALTER DATABASE BACKUP CONTROLFILE to 'location'; or use Recovery Manager (RMAN).

Archived Redo Log Backups
Archived redo logs are the key to successful media recovery. Depending on the disk space available and the number of transactions executed on the database, you want to keep as many days of archive logs on disk and you want to back them up regularly to ensure a more complete recovery.

Configuration Files
Configuration files may consist of spfile or init.ora, password file, tnsnames.ora, and sqlnet.ora. Since these files do not change often, then they require a less frequent backup schedule. If you lost a configuration file it can be easily recreated manually. When restore time is a premium, it will be faster to restore a backup of the configuration file then manually creating a file with a specific format.

WHICH BACKUP METHOD SHOULD BE USED?
Oracle provides users a choice of several basic methods for making backups. The methods include: