|Oracle9i Backup and Recovery Concepts
Release 1 (9.0.1)
Part Number A90133-02
The action that occurs when the redo log entry marking the checkpoint changes. For example, the CKPT process may record one record as the checkpoint, then three seconds later record a later log entry as the checkpoint. This action moves the checkpoint forward by saving all changes to the datafiles before the SCN reflected by the new checkpoint. Advancing the checkpoint reduces the amount of data that potentially requires recovery.
A copy of one of the filled members of an online redo log group made when the database is in
ARCHIVELOG mode. After the LGWR process fills each online redo log with redo records, the archiver process copies the log to one or more offline archive log destinations. This copy is the archived redo log.
The mode of the database in which Oracle copies filled online redo logs to disk. Specify the mode at database creation or by using the
DATABASE statement. You can enable automatic archiving either dynamically using the
SYSTEM statement or by setting the initialization parameter
Running the database in
ARCHIVELOG mode has several advantages over
NOARCHIVELOG mode. You can:
To protect the
ARCHIVELOG mode database in case of failure, back up the archived logs.
The operation in which the archiver background process copies filled online redo logs to offline destinations. An offline copy of an online redo logs is called an archived redo log. You must run the database in
ARCHIVELOG mode to archive redo logs.
A unit that contains one or more tape drives, a robotic arm, and a shelf of tapes. The ATL, also called a tape silo, is able to load and unload tapes into the tape drive from the shelf without operator intervention. More sophisticated tape libraries are able to identify each tape; for example, the robotic arm can use a bar-code reader to scan each tape's barcode and identify it.
See Also: media manager
The persistent preconfiguration of RMAN channels. You can use the
CONFIGURE command to specify disk and tape channels. Then, you can issue commands such as
RESTORE at the RMAN command prompt without manually allocating channels. RMAN uses whatever preallocated channels that it needs in order to execute the commands.
A mode of the database in which undo data is stored in a dedicated undo tablespace. Unlike in manual undo management mode, the only undo management that you must perform is the creation of the undo tablespace. All other undo management is performed automatically.
(1) A database created from target database backups using the RMAN
(2) A temporary database that is restored to a new location and then started up with a new instance name during tablespace point-in-time recovery (TSPITR). A TSPITR auxiliary database contains the recovery set and auxiliary set.
In TSPITR, the set of files that is not in the recovery set but which must be restored in the clone database for the TSPITR set to be successful. These auxiliary files include:
(1) A copy of data, that is, a database, tablespace, table, datafile, control file, or archived redo log. You can make a backup by:
(2) An RMAN command that creates a backup set.
See closed backup
The set of concepts, procedures, and strategies involved in protecting the database against data loss due to media failure or users errors. In a wider sense, backup and recovery also involves performing maintenance on backups as well as keeping records.
A backup of the control file. Make the backup by:
copycommand. Never create a backup control file by using operating system commands
Typically, you restore backup control files when all copies of the current control file are damaged; sometimes you restore them before performing certain types of point-in-time recovery.
See Also: control file
The database mode (also called hot backup mode) initiated when you issue the
BACKUP command before taking an online backup. You take a tablespace out of backup mode when you issue the
You must use this command when you make an operating system backup of one or more datafiles in an online tablespace. Recovery Manager does not require you to put the database in backup mode. Updates to tablespaces in backup mode create more than the usual amount of redo because each change causes Oracle to write the entire block rather than just the changed data to the redo log.
A backup piece is a physical file in an RMAN-specific format that belongs to only one backup set. A backup set usually contains only one backup piece. The only time RMAN creates more than one backup piece in a backup set is when you limit the backup piece size using the
MAXPIECESIZE option of the
The number of backups of a given file.
See retention policy
A backup of one or more datafiles, control files, or archived logs produced by the RMAN
BACKUP command. A backup set is a logical grouping of one or more binary files called backup pieces. Backup sets are in a proprietary format and can only be restored by RMAN.
The recovery of specified blocks within a datafile by using the Recovery Manager
BLOCKRECOVER command. Block media recovery leaves the affected datafiles online and restores and recovers only the damaged or corrupted blocks.
The termination of a disk mirroring procedure so that a mirror image contains a static copy of the current data. The broken mirror is no longer kept up-do-date. You can create operating system database backups by placing the tablespaces in the database in backup mode and then breaking the mirror. After taking the tablespaces out of backup mode, back up the broken mirror side to tape. After the backup is complete, you can resilver the mirror.
See Also: resilvering a mirror
The portion of the SGA that holds copies of Oracle data blocks. All user processes concurrently connected to the instance share access to the buffer cache.
The buffers in the cache are organized in two lists: the dirty list and the least recently used (LRU) list. The dirty list holds dirty buffers, which contain data that has been modified but has not yet been written to disk. The least recently used (LRU) list holds free buffers (unmodified and available), pinned buffers (currently being accessed), and dirty buffers that have not yet been moved to the dirty list.
See Also: SGA (System Global Area)
There is a delay between when a data block is changed in the buffer cache and when it is saved to the datafiles on disk. Hence, an Oracle instance may crash before a data block has been saved. To prevent loss of these changes, Oracle writes the change to the data block (and the change to the rollback or undo blocks) to the online redo log before making the change to the block in the cache.
In cache recovery, Oracle rolls forward to recover data that has not been recorded in the datafiles. Unlike media recovery, crash recovery never needs to read the contents of any archived logs: all of the changes needed are in the online redo logs.
A type of incomplete media recovery in which you use the
RECOVER command with the
CANCEL clause. Recovery proceeds until you issue the
The description of a change, usually to a single Oracle block. A set of change vectors form the content of a redo record. A change vector is the smallest unit of change recorded in the redo log.
See Also: redo record
A type of incomplete media recovery that recovers up to a specified SCN. You can also perform cancel-based recovery, which recovers until you issue the
CANCEL command, and time-based recovery, which recovers to a specified time.
A connection between RMAN and the target database. Each allocated channel starts a new Oracle server session; the session then performs backup, restore, and recovery operations. The type of channel determines whether the Oracle server process will attempt to read or write and whether it will work through a third-party media manager. If the channel is of type:
DISK, the server process reads backups from or write backups to disk
sbt, the server process reads backups from or write backups to a third-party media manager
Channels are always able to read and write datafiles to and from disk, no matter what their type.
RMAN parameters that allow you to control I/O for backups and copies created by a channel.
A data structure that defines an SCN in the redo thread of a database. Checkpoints are recorded in the control file and each datafile header, and are a crucial element of recovery. The three types of checkpoints are: thread checkpoint, database checkpoint, and datafile checkpoint.
A numeric value that is mathematically derived from the contents of an Oracle data block. The checksum allows Oracle to validate the consistency of the block.
See Also: data block
Control file records containing non-critical information used by RMAN for backups and recovery operations. These records are arranged in a logical ring. When all available record slots are full, Oracle either expands the control file to make room for a new records or overwrites the oldest record. The
CONTROL_FILE_RECORD_KEEP_TIME initialization parameter controls how long a given record must be kept before it can be overwritten.
See Also: noncircular reuse records
A database shut down with the
NORMAL options of the
SHUTDOWN statement. A database shut down cleanly does not require recovery; it is already in a consistent state.
A backup of one or more database files taken while the database is closed. Typically, closed backups are also whole database backups. If you closed the database cleanly, then all the files in the backup are consistent. If you shut down the database using a
ABORT or the instance terminated abnormally, then the backups are inconsistent.
A database that is not available to users for queries and updates. When the database is closed you can start the instance and optionally mount the database.
See Also: open database
Multiple nodes, each of which is capable of accessing data on a set of shared disks. In a cold failover cluster, the database instance is active on only one node. If the instance on the active node crashes, a script can automatically start an instance on the passive node and recover the database. In an Oracle Real Applications Cluster configuration, each node has an active instance against the same database and can perform I/O on a shared disk at the same time.
See closed backup
A file containing a sequence of RMAN commands that you can run from the command line. The contents of the command file should be identical to commands entered at the command line.
Recovery of one or more datafiles that applies all online and archived redo generated after the restored backup. Typically, you perform complete media recovery when media failure damages one or more datafiles or control files. You fully recover the damaged files using all redo generated since the restored backup was taken. If you use RMAN, you can also apply incremental backups during complete recovery.
The process of copying only used data blocks into RMAN backup sets. A newly created datafile contains many never-used blocks. When RMAN creates backup sets, it only includes blocks that have been used; it follows that RMAN does not write never-used blocks into backup sets.
A whole database backup that you can open with the
RESETLOGS option without performing media recovery. In other words, you do not need to apply redo to datafiles in this backup for it to be consistent. All datafiles in a consistent backup must:
You can only take consistent backups after you have made a clean shutdown of the database. The database must not be opened until the backup has completed.
A binary file associated with a database that maintains the physical structure and time stamps of all files in that database. Oracle updates the control file continuously during database use and must have it available for writing whenever the database is mounted or open.
The automatic backup of the current control file that RMAN makes in the situations:
COPYcommand run at the RMAN prompt
COPYcommand within a
RUNblock that is not followed by another
The control file autobackup has a default filename that allows RMAN to restore it even if the control file and recovery catalog are lost. You can override the default filename if desired.
(1) To replicate data. You make copies of Oracle datafiles, control files, and archived redo logs in two ways:
(2) A Recovery Manager command that makes a replica of a database's datafiles, control file, or archived redo logs. This replica is made by an Oracle server process, allocated to a Recovery Manager channel, which reads the Oracle file and writes a replica out to disk. Recovery Manager can copy the files of an open database without putting the tablespaces into backup mode.
See Also: backup
In RMAN, the datafile in the target database pointed to by the control file. You can make a datafile copy current again by executing a
An Oracle block that is not in a recognized Oracle format, or whose contents are not internally consistent. Typically, corruptions are caused by faulty hardware or operating system problems. Oracle identifies corrupt blocks as one of two types:
You can only repair a media corrupt block by:
If media corruption is due to faulty hardware, neither solution will work until the hardware fault is corrected.
A datafile that contains one or more corrupt blocks.
See Also: corrupt block
The automatic application of online redo records to a database after either a single-instance database crashes or all instances of an Oracle Real Applications Cluster configuration crash. Crash recovery only requires redo from the online logs: archived redo logs are not required.
In crash recovery, an instance automatically recovers the database before opening it. In general, the first instance to open the database after a crash or
ABORT automatically performs crash recovery.
A check to determine whether files on disk or in the media management catalog correspond to the information in the recovery catalog (if used) and the control file. Because the media manager can mark tapes as expired or unusable, and because files can be deleted from disk or otherwise become corrupted, the recovery catalog and control file can contain outdated information about backups and image copies.
CROSSCHECK command to perform a crosscheck. To determine whether you can restore a file, run
See Also: validation
An incremental backup that backs up all the blocks changed since the most recent backup at level n-1 or lower. For example, in a cumulative level 2 backup, RMAN determines which level 1 or level 0 backup is most recent and then backs up all blocks changed since that backup. When recovering with cumulative incremental backups, only one backup at each level needs to be applied.
The control file on disk; it is the most recently modified control file for the current incarnation of the database. For a control file to be considered current during recovery, it must not have been restored from backup.
The online redo log file in which the LGWR background process is currently logging redo records. Those files to which LGWR is not writing are called inactive.
Every database must contain at least two online redo log files. If you are multiplexing the online redo logs, LGWR concurrently writes the same redo data to multiple files. The individual files are called members of an online redo log group.
The smallest unit of data in an Oracle database. Every database has a default block size, although data blocks in different tablespaces can have different sizes.
The location identifier of an Oracle data block. A data block address is constituted by a datafile number and a data block number. You can specify the
DBA of a data block in the
The number that identifies a specific data block within a datafile. Blocks in a datafile are numbered sequentially. You can specify the data block number within the
The thread checkpoint that has the lowest SCN. The database checkpoint guarantees that all changes in all enabled threads prior to the database checkpoint have been written to disk.
See Also: checkpoint
The recovery of a database to a specified noncurrent time, SCN, or log sequence number.
A datafile is a physical operating system file on disk that was created by Oracle and contains data structures such as tables and indexes. A datafile can only belong to one database.
See Also: inaccessible datafile
A copy of a datafile on disk produced by either:
The checkpoint structure stored in the header of each datafile. All redo in all threads prior to the datafile checkpoint SCN is guaranteed to have been saved to the datafile.
The application of redo records to a restored datafile in order to roll it forward to a more current time. Unless you are doing block media recovery, the datafile must be offline while being recovery.
An internal, uniquely generated number that differentiates databases. Oracle creates this number automatically when you create the database.
When using the Oracle Managed Files feature, the default directory specifies where Oracle creates files when no file specification has been given in the creation operation. You define the location through the initialization parameter
A type of incremental backup that backs up all blocks that have changed since the most recent backup at level
n or lower. For example, in a differential level 2 backup RMAN determines which level 2, level 1, or level 0 backup is most recent and then backs up all blocks changed since that backup. Differential backups, also called non-cumulative incremental backups, are the default type of incremental backup. When recovering using differential incremental backups, one or more backups at each level must be applied.
A buffer in the database buffer cache that contains a change that has not yet been written to the datafiles.
A hardware component that is responsible for controlling one or more disk drives. The term applies to controllers integrated with the disk drive they control, as well as to high performance disk array controllers supporting various RAID configurations.
A database created from target database backups using the RMAN duplicate command.
See Also: auxiliary database
The extraction of logical data (that is, not physical files) from a database using the Export utility. You can then use the Import utility to import the data into a database.
The first block of an Oracle datafile. The file header contains bookkeeping information related to the file, including the checkpoint SCN. Oracle requires media recovery when the checkpoint SCN in the datafile header does not match the file header information stored in the control file.
See Also: thread checkpoint
A software package that manipulates file systems.
See Also: file system
A file system is a data structure built inside a contiguous disk address space. One computer can have multiple file systems, each independent of the others.
The file system allows a hard disk to contain files. Each file on the file system is distinguished by a unique filename. A file system is commonly built on top of a logical volume constructed by a logical volume manager (LVM).
A type of media corruption that can occur when database writer is writing a block at the same time an operating system utility is reading the block for backup. The block that the operating system reads can be split, that is, the top of the block is written at one point in time while the bottom of the block is written at another point in time. If you restore a file containing a fractured block and Oracle reads the block, then the block is considered a corrupt block.
The potential for fractured blocks necessitates putting tablespaces in backup mode before user-managed online backups. A database in backup mode writes whole Oracle data blocks to the redo log, so that if a block is split during the backup, you can repair it by using redo. Recovery Manager does not experience this problem because the server process performing the backup or copy reads each block to determine whether it is split and re-reads the block until it gets a consistent version.
See Also: corrupt datafile
A non-incremental RMAN backup. Note that "full" does not refer to how much of the database is backed up, but to the fact that the backup is not incremental. Consequently, you can make a full backup of one datafile.
The only difference between a full backup and an incremental level 0 backup is that the full backup will not affect the number of blocks backed up by any subsequent incremental backup.
An export of the whole database.
A Recovery Manager operation that updates the recovery catalog with all changed information in the database's control file. You can initiate a full catalog resynchronization by issuing the RMAN command
CATALOG. Recovery Manager initiates resynchronization operations as needed when executing certain commands.
A datafile that contains at least one block with an SCN more recent than the checkpoint SCN in its header. For example, this situation occurs when Oracle updates a datafile that is in backup mode. A fuzzy file that is restored always requires recovery.
See Also: thread checkpoint
See online backup
See backup mode
A copy of a single datafile, archived redo log file, or control file that is:
COPYcommand or an operating system command such as the UNIX
A datafile that Oracle is attempting to read, but cannot find. Attempts to access an inaccessible file result in errors. Typically, a file is inaccessible because the media on which it is stored is faulty or the file has been moved or deleted.
See Also: media failure
A redo log file that is not required for crash or instance recovery because the changes contained in its redo records have already been applied to the database. The current online redo log is never inactive. If you operate the database in
ARCHIVELOG mode, the archiver process archives inactive redo log files.
A separate version of a physical database. The incarnation of the database changes when you open it with the
RESETLOGS option. Make a whole database backup of all files that are not offline-clean or read-only after opening with the
RESETLOGS option. Note that if you run the RMAN command
RESETLOGS, RMAN automatically resets the database incarnation.
The recovery of a database in which you do not apply all of the changes generated since you created the restored backup.
Incomplete recovery is usually performed when:
The requirement is to recover up until some point in time before an incorrect action occurred in the database. For example, a user mistakenly deletes payroll transactions before the transactions are sent to the payroll agency. In this example, the DBA will need to restore the whole database and then perform incomplete recovery up until the point just before the user deleted the transactions.
An archived redo log which is needed for complete recovery was not backed up, or the archived redo log contents are corrupt. In this case, you only option is to recover up to the missing log.
In each case, open the database with the
RESETLOGS option after performing media recovery.
A backup in which some of the files in the backup contain changes that were made after the files were checkpointed. This type of backup needs recovery before it can be made consistent. Inconsistent backups are usually created by taking open database backups; that is, the database is open while the files are being backed up. You can also make an inconsistent backup by backing up datafiles while a database is closed, either:
Note that inconsistent backups are only useful if the database is in
An RMAN backup in which only modified blocks are backed up. Incremental backups are classified by level. An incremental level 0 backup performs the same function as a full backup in that they both back up all blocks that have ever been used. The difference is that a full backup will not affect blocks backed up by subsequent incremental backups, whereas an incremental backup will affect blocks backed up by subsequent incremental backups.
Incremental backups at levels greater than 0 back up only blocks that have changed since previous incremental backups. Blocks that have not changed are not backed up. An incremental backup can be either a differential incremental backup or a cumulative incremental backup.
An SGA (System Global Area), Oracle code, and background processes. Create an instance by issuing any of the following commands:
The instance starts, but does not mount the control file or open the database.
The instance mounts the control file but does not open the database.
The instance starts, mounts the control file, and opens the database.
An instance is terminated by executing a
The termination of an Oracle instance due to a hardware failure, application error, or
ABORT statement. Strictly speaking, an instance failure occurs whenever the database is not shut down cleanly (that is, with a
TRANSACTIONAL statement). Crash or instance recovery is always required after an instance failure.
In an Oracle Real Applications Cluster configuration, the application of redo data to an open database by an instance when this instance discovers that another instance has crashed. A surviving instance automatically uses the redo log to recover the data in the instance's buffer cache. Oracle undoes any uncommitted transactions that were in progress on the failed instance when it crashed and then clears any locks held by the crashed instance after recovery is complete.
The contents of an RMAN
See Also: job commands
RMAN commands such as
RECOVER that you can execute at the RMAN prompt or within the brackets of a
See Also: standalone commands
A utility that allows log files to be read, analyzed, and interpreted by means of SQL statements. LogMiner can view any valid online or archived redo log from Oracle8 and higher databases. You can use LogMiner to do the following:
See Also: archived redo log
A number that uniquely identifies a set of redo records in a redo log file. When Oracle fills one online redo log file and switches to a different one, Oracle automatically assigns the new file a log sequence number. For example, if you create a database with two online log files, then the first file is assigned log sequence number 1. When the first file fills and Oracle switches to the second file, it assigns log sequence number 2; when it switches back to the first file, it assigns log sequence number 3, and so forth.
For RMAN, a type of incomplete recovery that recovers up to a specified log sequence number.
See Also: incomplete recovery
The point at which LGWR stops writing to the active redo log file and switches to the next available redo log file. LGWR switches when either the active log file is filled with redo records or you force a switch manually.
If you run the database in
ARCHIVELOG mode, Oracle archives the redo data in inactive log files into archived redo logs. When a log switch occurs and LGWR begins overwriting the old redo data, you are protected against data loss because the archived redo log contains the old data. If you run in
NOARCHIVELOG mode, Oracle overwrites old redo data at a log switch without archiving it. Hence, you lose all old redo data.
See Also: redo log
Backups in which the Export utility uses SQL to read database data and then export it into a binary file at the operating system level. You can then import the data back into a database using the Import utility.
Backups taken with the Export utility differ in the following ways from RMAN backups:
See Also: physical backups
A software program that allows sections of multiple physical disks to be combined into a single contiguous address space. This space appears as one disk to higher layers of software.
A backup that you want to exclude from an expiration policy, but want to record in the recovery catalog. Typically, long-term backups are snapshots of the database that you may want to use in the future for report generation. For example, you may want to survey employee salaries in past years.
A mode of a standby database in which the standby waits for archived log files from a target database and then automatically applies the redo logs once the files become available. This feature eliminates the need for you to interactively provide the recovery process with filenames of the archived redo logs.
A mode of the database in which undo blocks are stored in user-managed rollback segments. In automatic undo management mode, undo blocks are stored in a system-managed, dedicated undo tablespaces.
The desired time required to perform instance or media recovery on the database. For example, you may set 10 minutes as the goal for media recovery from a disk failure. A variety of factors influence MTTR for media recovery, including the speed of detection, the type of method used to perform media recovery, and the size of the database.
A physical problem that arises when Oracle fails in its attempt 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 the datafiles, redo log files, and control files. Because the database instance cannot continue to function properly, it cannot write the data in the buffer cache of the SGA to the datafiles.
See Also: media recovery
A utility provided by a third party vendor that is capable of actions such as loading, labelling and unloading sequential media such as tape drives. Media managers also allow you to configure media expiration and recycling, and may also have the ability to control an ATL (automated tape library).
An Oracle published API to which media management vendors have written compatible software libraries. This software integrates with Oracle so that an Oracle server process is able to issue commands to the media manager to write backup files to sequential storage, and read files from sequential storage. When Oracle issues a request to backup or restore a file, the media manager handles the actions required to load, label, and unload the correct tape.
The media management interface is also called the media management layer, the media management library (MML), and the SBT interface.
The application of redo or incremental backups to a restored backup datafile or individual data block to bring it to a specified time. Datafile media recovery always begins at the lowest SCN recorded in the datafile headers.
When performing media recovery, you can recover:
If you use all redo data, you perform complete recovery; if you use only part of the redo data, you perform incomplete recovery. Typically, you perform media recovery after a media failure damages some or all of the database files (datafiles, control files, or online redo logs).
ARCHIVELOG mode, you have the choice of complete recovery or incomplete recovery. In
NOARCHIVELOG mode, the only option is typically to restore from the most recent backup without applying redo data.
Maintaining identical copies of data on one or more disks. Typically, mirroring is performed on duplicate hard disks at the operating system level, so that if one of the disks becomes unavailable, the other disk can continue to service requests without interruptions. For example, you can mirror a datafile so that Oracle writes the same information to two different disk drives. The operation of breaking a mirror splits off the copy and makes a backup, while resilvering a mirror rejoins the split copy.
When mirroring files, Oracle writes once while the operating system writes to multiple disks; when multiplexing files, Oracle writes the same data to multiple files.
An instance that is started and has the control file associated with the database open. You can mount a database without opening it; typically, you put the database in this state for maintenance or for restore and recovery operations.
RMAN-generated incremental backups that allow you to conserve space by planning which blocks to back up and when. A level 0 incremental backup, which is the base for subsequent incremental backups, copies all blocks containing data. When you generate a level
n incremental backup in which
n is greater than
0, you back up either:
nor lower. This is the default type of incremental backup, called a differential incremental backup.
-1or lower. This type of backup is called a cumulative incremental backup.
You can create a backup strategy in which you generate a backup at a different level each day, thereby controlling how much data you back up.
The automated maintenance of more than one identical copy of the online redo log. To multiplex the online logs, create multiple members in each redo log group. The degree of multiplexing is directly related to the number of members in each group.
The automated maintenance of more than one identical copy of a database's control file. To multiplex the control file, create multiple entries in the
CONTROL_FILES initialization parameter.
The RMAN technique of reading database files simultaneously from the disks and then writing the blocks to the same backup piece. The degree of multiplexing is smaller of these two parameter settings:
BACKUP command) and
CHANNEL). The degree of multiplexing can be smaller than number of files included in the backup piece. For example, if RMAN writes ten datafiles into a backup set but reads from only one at a time (because
1), then blocks from ten datafiles are included in the backup piece but the degree of multiplexing is one.
The Oracle archiver process is able to archive multiple copies of a redo log. You can multiplex archived redo logs by setting
n is an integer) in the initialization parameter file.
See Also: mirroring
Using multiple archiver processes (ARC
n) to archive online redo logs to one or more locations. Multiple archiver processing prevents the bottleneck that occurs when LGWR writes to the online redo log faster than a single archive process can write to the archive destinations. You can enable this feature at startup or at runtime by setting the initialization parameter
n is any integer from 1 to 10.
The mode of the database in which Oracle does not require filled online redo logs to be archived before they can be overwritten. Specify the mode at database creation or change it by using the
DATABASE command. Oracle does not recommend running in
NOARCHIVELOG mode because it severely limits the possibilities for recovery of lost data.
During block media recovery, redo records are applied serially to a block. When a block is newed all previous redo for the block becomes irrelevant because the redo applies to a defunct incarnation of the block. Because blocks can be newed, block media recovery can sometimes complete successfully even if redo records are missing.
See Also: corrupt block
Control file records containing critical information needed by the Oracle database. These records do not change often and cannot be overwritten. Some examples of information in non-circular reuse records include:
See Also: circular reuse records
The transmittal of archived redo log files to a local disk.
See Also: standby transmission
An RMAN backup or image copy is obsolete when it is no longer needed for media recovery, for example, when multiple more recent backups and copies exist. A retention policy determines when a backup or copy is obsolete.
A backup of a tablespace or datafile made when the tablespace or datafile is offline and the database open. Run the
OFFLINE statement to take a tablespace offline, and the
OFFLINE statement to take an individual datafile offline.
When a tablespace is taken offline clean, it is taken offline using the
NORMAL statement. The datafiles in the tablespace are checkpointed and do not require recovery before being brought online. If a tablespace is not taken offline clean, then its datafiles must be recovered before being brought online.
The SCN that specifies when a datafile was brought online after being offline, or made read/write after being read-only. This SCN is stored in the control file and is the last SCN in the offline range. The offline-end checkpoint is important because it indicates that changes after this SCN are required to recover the datafile.
The span between the offline-start SCN and offline-end checkpoint fields of the record for a datafile in the control file. The offline range specifies a period during which there is guaranteed to be no redo for the datafile, because during this range the datafile was offline-normal or read-only. Thus, media recovery can skip this log range when recovering the datafile.
The SCN that specifies when a datafile was taken offline cleanly or made read-only. This SCN is stored in the control file and is the first SCN in the offline range. The offline-start SCN is important for recovery because it indicates that no changes made between this SCN and the offline-end checkpoint are required to recover the datafile.
A tablespace that is not available to users when the database is open. You can only take a tablespace offline while the database is open. If a tablespace is taken offline, all online datafiles contained in the tablespace are taken offline.
You can take a tablespace offline using the
OFFLINE statement with three different options:
All the files in the tablespace are checkpointed, then taken offline. If any datafile belonging to the tablespace is not available, the tablespace cannot be taken offline normal. Datafiles in a tablespace taken offline cleanly do not need to be recovered before the tablespace is brought back online.
All datafiles in the tablespace that are accessible to Oracle are checkpointed, then taken offline. Files that were checkpointed by the
TEMPORARY command do not need recovery. Datafiles that were not checkpointed because they were not accessible at the time of an
IMMEDIATE command must be recovered before the tablespace is brought back online.
All files in the tablespace are taken offline without any attempt to checkpoint the files first. All files in the tablespace must be recovered before the tablespace is brought online.
See Also: offline datafile
A datafile that is not available to users when the database is open. In exceptional circumstances, Oracle will automatically take a datafile offline if required. This file will need recovery before it can be brought online.
You can take a datafile offline either:
OFFLINEwhen the database is mounted or open.
If you take an individual datafile offline, then you must recover it before bringing it back online.
See Also: offline tablespace
A backup of one or more datafiles taken while a database is open and the datafiles are online. When you make a user-managed backup while the database is open, you must put the tablespaces in backup mode by issuing an
BACKUP command. When you make an RMAN backup while the database is open, however, you do not need to put the tablespaces in backup mode.
A datafile that users can access. The database can be open or mounted when you issue the command
ONLINE. If the database is open, the datafile must be consistent with the rest of the database before you can bring it online. If the database is mounted, then you can bring the datafile online without being consistent with the other datafiles, but it will require recovery before the database is opened.
See Also: online tablespace
The online redo log is a set of two or more files that record all changes made to Oracle datafiles and control files. Whenever a change is made to the database, Oracle generates a redo record in the redo buffer. The LGWR process flushes the contents of the redo buffer into the online redo log.
The current online redo log is the one being written to by LGWR. When LGWR gets to the end of the file, it performs a log switch and begins writing to a new log file. If you run the database in
ARCHIVELOG mode, then the archiver process or processes copy the redo data into an archived redo log.
See Also: archived redo log
The Oracle online redo log consists of two or more online redo log groups. Each group contains one or more identical online redo log members. An online redo log member is a physical file on the operating system containing the redo records.
A physical online redo log file within an online redo log group. Each log group must have one or more members. Each member of a group is identical.
A tablespace that is available to users while the database is open. You can make a tablespace available for access by users by issuing the command
ONLINE. The database must be open to alter a tablespace online, and all files in the tablespace must be consistent with the rest of the database before the tablespace can be made online.
See Also: online datafile
A database that is available to users to query and update. The database is opened either automatically through a
STARTUP statement or explicitly through an
A file that is created automatically by the Oracle database server when it is needed and automatically deleted when it is no longer needed.
Backups and copies that are unusable because they belong to incarnations of the database that are not direct ancestors of the current incarnation.
A form of recovery in which several processes simultaneously apply changes from redo log files. Instance and media recovery can be parallelized automatically by specifying the
RECOVERY_PARALLELISM initialization parameter or options to the SQL/SQL*Plus
RECOVER command. Oracle uses one process to read the log files sequentially and dispatch redo information to several recovery processes, which apply the changes from the log files to the datafiles.
See Also: serial recovery
Allocating multiple channels for Recovery Manager backup and recovery operations. You can parallelize:
A type of resynchronization in which RMAN transfers information about archived redo logs, backup sets and datafile copies from the target database control file to the recovery catalog. Partial resynchronization does not transfer information such as:
A file created by the
ORAPWD command. A database must use password files if you wish to connect using the
SYSOPER roles over a network. For a more comprehensive explanation, see the Oracle9i Database Administrator's Guide.
Physical database files that have been copied from one place to another. The files can be datafiles, archived redo logs, or control files. You can make physical backups using Recovery Manager or with operating system commands such as the UNIX
The datafiles, control files, and redo logs in a database at a given time. Issue the RMAN
SCHEMA command to obtain a list of tablespaces and datafiles.
A full resynchronization of the recovery catalog updates all changed RMAN metadata in the repository, including physical schema information. If the database is open, RMAN also gathers information about rollback segments. A partial resynchronization of the recovery catalog does not update physical schema or rollback information.
The earliest time within a recovery window. A retention policy that specifies a recovery window mandates that the database must be able to be recovered to the earliest time in the window. For example, in a recovery window of 7 days, the point of recoverability is always 7 days before the current time.
See Also: obsolete backups and copies
The functionality that enables a media manager to take over the transfer of data between the media storage device and disk during RMAN backup and restore operations.
Errors that occur when Oracle is unable to read a datafile, control file, or online redo log. Oracle returns an error to the operating system and to the application, along with an Oracle error indicating that the file cannot be found, cannot be opened, or cannot be read. Note that Oracle does not automatically take a datafile offline if it is unable to read it.
A database opened with the
ONLY command. As their name suggests, read-only databases are for queries only and cannot be modified. Oracle allows a standby database to be run in read-only mode, which means that it can be queried while still serving as an up-to-date emergency replacement for the primary database.
A tablespace whose status has been changed to prevent it from being updated. You put in read-only mode by executing the SQL statement
ONLY. Typically, you put a tablespace in read-only mode to reduce the frequency with which it is backed up. For example, instead of backing up the tablespace nightly, you reduce the backup frequency to once a month.
(1) A Recovery Manager command that updates a restored datafile by the application of incremental backups (if they exist) and then by the application of archived or online redo logs.
(2) A SQL*Plus command that updates a restored file by the application of archived or online redo logs.
See Also: recovery
The application of redo data or incremental backups to database files in order to reconstruct lost changes. The three types of recovery are instance recovery, crash recovery, and media recovery. Oracle performs the first two types of recovery automatically using online redo records; only media recovery requires you to restore a backup and issue commands. Only Recovery Manager can recover datafiles by applying incremental backups.
A set of Oracle tables and views used by Recovery Manager to store information about Oracle databases. Recovery Manager uses this data to manage the backup, restore, and recovery of Oracle databases. The recovery catalog is optional. If you choose not to use a recovery catalog, RMAN uses the target database control file as the sole repository of metadata.
See Also: recovery catalog database
An Oracle database that contains a recovery catalog schema. You should not store the recovery catalog in the target database.
A utility that backs up, restores, and recovers Oracle databases. You can use it with or without the central information repository called a recovery catalog. If you do not use a recovery catalog, RMAN uses the database's control file to store information necessary for backup and recovery operations. You can use RMAN in conjunction with a media manager to back up files to tertiary storage.
A environment containing the Recovery Manager (RMAN) executable and the various computers, databases, applications, and APIs with which it interacts.
One or more tablespaces that are being recovered to an earlier point in time during tablespace point-in-time recovery (TSPITR). After TSPITR, all database objects in the recovery set have been recovered to the same point in time.
See Also: auxiliary set
A recovery window is a period of time in a retention policy bounded by the current time and the earliest point of recoverability. The point of recoverability is the end time for a hypothetical point-in-time recovery, that is, the point to which you must be able to recover following a media failure. A retention policy states that you must have enough backups and archived redo logs to be able to recover to any point between the current time and the point of recoverability.
For example, if you implement a recovery window of one week, then this window of time must extend back exactly seven days from the present so that you can restore a backup and recover it to any point within the last week.
The recovery window always keeps pace with the current time. For example, if the current day is January 14 and the recovery window is 7 days, then the recovery window stretches between January 7 and January 14. When the current day is January 28, then the recovery window stretches between January 21 and January 28.
Any backups or logs not needed for recovery are considered obsolete. If you make periodic backups, then as the recovery window moves forward in time the older backups become obsolete.
A redo log can be either an online redo log or an archived redo log. The online redo log is a set of two or more redo log groups that records all changes made to Oracle datafiles and control files. An archived redo log is a copy of an online redo log that has been copied to an offline destination. If the database is in
ARCHIVELOG mode and automatic archiving is enabled, then the archiver process or processes copy each online redo log to one or more archive log destinations after it is filled.
The memory buffer in the system global area (SGA) in which Oracle logs redo records. The background process LGWR writes the buffers into the current online redo log.
See Also: redo record
Redo log files are the operating system files that log writer writes its changes. A redo log member within a redo log group corresponds to one and only one redo log file. The
V$LOGFILE view displays the filenames of redo log files.
Each online redo log member (which corresponds to an online redo log file) belongs to a group. A group has one or more identical members. A multiplexed redo log is a redo log in which the redo groups have multiple members.
A group of change vectors describing a single, atomic change to the database. Oracle constructs redo records for all data block changes and saves them on disk in the current online redo log. Redo records allow changes to database blocks to be reconstructed should data loss occur.
See Also: redo log
The redo generated by an instance. If the database runs in a single instance configuration, then the database has only one thread of redo. If you run in an Oracle Real Application Clusters configuration, then you have multiple redo threads, one for each instance.
A set of backups enabling you to recover from the failure or loss of any Oracle database file.
In RMAN, the execution of a
DATABASE command in order to record the existence of a target database in the recovery catalog. A target database is uniquely identified in the catalog by its DBID. You can register more than one database in the same catalog, and also register the same database in multiple catalogs.
See Also: DBID
The collection of RMAN metadata about backup and recovery operations on the target database. Either the control file or the recovery catalog can function as the RMAN repository.
A method for opening a database that results in a new database incarnation, the resetting of the log sequence number to 1, and the re-formatting or re-creation of the online redo logs. A database must be opened with the
RESETLOGS keyword after:
Together with the
RESETLOGS time stamp, the
RESETLOGS SCN uniquely identifies each execution of an
RESETLOGS statement. When the online logs are reset, Oracle creates a new and unique incarnation of the database.
RESETLOGS SCN and time stamp are stored in the control file, in each datafile header, and in each redo log file header. An online or archived redo log cannot be applied by recovery if its
RESETLOGS data does not match the database information in the control file.
Except for special circumstances (for example, offline normal or read-only tablespaces), a datafile cannot be recovered or accessed if its
RESETLOGS SCN and time stamp do not match the database information in the control file. This precaution ensures that changes discarded by
RESETLOGS cannot get reapplied to the database.
Configuring the operating system or hardware managing the mirror so that you refresh a broken mirror from the half that is up-to-date and then maintain both sides of the mirror.
The replacement of a lost or damaged file with a backup. You can restore files either with operating system commands such as UNIX
cp or the RMAN
See Also: recover
The operation that updates the recovery catalog with current information from the target database control file. You can initiate a full resynchronization of the catalog by issuing a
CATALOG command. A partial resynchronization transfers information to the recovery catalog about archived redo logs, backup sets and datafile copies. RMAN performs resynchronizations automatically when needed.
A user-defined policy for determining how long backups and copies need to be retained for media recovery. You can define a retention policy in terms of backup redundancy or a recovery window. For example, if you need to recover the database to any point within the last 7 days, then at least one backup made 8 days ago or earlier must be retained. Also, you must retain all of the archived redo logs needed to perform the recovery.
Any backup or copy that is not needed according to the retention policy is considered obsolete. RMAN can automatically delete such obsolete backups.
See Also: obsolete backups and copies
Database segments that record the before-images of changes to the database. A rollback segment contains a transaction table with two or more extents of undo blocks. Undo blocks are arranged in a circular fashion so that older pieces of undo data are overwritten in chronological order.
The application of redo records or incremental backups to datafiles and control files in order to recover changes to those files.
System Backup to Tape
See Also: media management interface
A form of recovery in which a single process applies the changes in the redo log files sequentially.
See Also: parallel recovery
A group of shared memory structures that contain data and control information for one Oracle database instance. The SGA and Oracle processes constitute an Oracle instance. Oracle automatically allocates memory for an SGA whenever you start an instance and the operating system reclaims the memory when you shut down the instance. Each instance has one and only one SGA.
A copy of a database's control file taken by Recovery Manager. RMAN uses the snapshot control file to read a consistent version of a control file when either resynchronizing the recovery catalog or backing up the control file.
See fractured block
The process of restoring archived logs from tertiary storage to disk in order to allow recovery to proceed. RMAN stages the logs to disk when the
RECOVER command is executed. To use this feature, you must configure a media manager.
See Also: media manager
RMAN commands that you do not have to execute within the brackets of a
See Also: job commands
A copy of a production database that you can use for disaster protection. You can update the standby database with archived redo logs from the production database in order to keep it current. If a disaster destroys the production database, you can activate the standby database and make it the new production database.
The transmittal of archived redo log files over a network to either a local or remote standby database.
Each control file record for a datafile has a field called the stop SCN. If the datafile is offline or read-only, the stop SCN is the SCN beyond which no further redo exists for the datafile. If the datafile is online and any instance has the database open, then the stop SCN is set to infinity. Oracle uses the stop SCN during media recovery to determine when recovery for a datafile can end. This feature ensures that media recovery terminates when recovering an offline file while the database is open.
A sequence of RMAN commands stored in the recovery catalog.
A Recovery Manager command which converts a datafile copy into a datafile used by an Oracle database. It performs the equivalent function of the SQL statement
', and also marks the datafile copy as no longer available.
A stamp that defines a committed version of a database at a point in time. Oracle assigns every committed transaction a unique SCN.
SYSTEM tablespace differs from other tablespaces in that all datafiles contained in the tablespace must be online for Oracle to function. If a media failure affects one of the datafiles in
SYSTEM, then you must mount the database and recover.
A database is divided into one or more logical storage units called tablespaces. Each tablespace has one or more physical datafiles exclusively associated with it.
See Also: datafile
The recovery of one or more non-
SYSTEM tablespaces to a point in time that is different from the database. You can use either RMAN or user-managed methods to perform TSPITR.
A user-specified character string that acts as a symbolic name for a backup set or image copy. You can specify a tag when executing the
CHANGE command. The maximum length of a tag is 30 characters.
The most recent redo record in the redo log file. As users make changes to the database, the tail keeps moving forward. Since the latest checkpoint is always temporally behind the tail of the log, it is said to lag the tail of the log. If the checkpoint lags the tail of the log significantly, crash recovery time increases.
Writing output to a tape drive fast enough to keep the tape constantly busy. If the device is not kept busy, then its performance decreases because the drive mechanism must be started and stopped for each piece of data that is received.
A piece of hardware that reads and writes magnetic tapes.
One physical piece of tape media.
In RMAN, the database that you are backing up or restoring.
A file that belongs to a temporary tablespace, and is created with the
TEMPFILE option. Temporary tablespaces cannot contain permanent database objects such as tables, and are typically used for sorting. Because tempfiles cannot contain permanent objects, RMAN does not back them up.
See redo thread
A type of checkpoint stored in the control file indicating that all changes to online datafiles in a given thread prior to the checkpoint SCN have been saved to disk. Oracle updates the thread checkpoint every time an instance checkpoints its thread.
Checkpointing is crucial for recovery because it limits the amount of transaction redo that crash and instance recovery must potentially apply. Online switch management guarantees that the current checkpoint has moved out of an online log file before that log file eligible for reuse. Checkpointing works in conjunction with online log switch management to ensure that crash and instance recovery can be accomplished using only online redo logs.
The incomplete recovery of database files to a noncurrent time. Time-based recovery is also known as point-in-time recovery. There are two types:
signifies the incomplete recovery of all datafiles and control file to a time before the most recent time
signifies the incomplete recovery of all datafiles in one or more tablespaces on an auxiliary database to a specific time before the most current time. The tablespace is then re-integrated into the original database.
Transaction recovery involves rolling back all uncommitted transactions of a failed instance. These are "in-progress" transactions that did not commit and that Oracle needs to roll back. It is possible for uncommitted transactions to get saved to disk. In this case, Oracle uses undo data to reverse the effects of any changes that were written to the datafiles but not yet committed.
A feature that transports a set of tablespaces from one database to another, or from one database to itself. Transporting or "plugging" a tablespace into a database is like creating a tablespace with preloaded data. This feature is often an advantage because:
Oracle blocks that contain the before-image of a change to the database. For example, if you update a salary column in a table from 55 to 65, Oracle writes the before-image of 55 into an undo block. If you run in manual undo management mode, then undo blocks are stored in user-managed rollback segments. If you run in automatic undo management mode, then undo blocks are stored in system-managed undo tablespaces.
A dedicated tablespaces that stores only undo information when the database is run in automatic undo management mode. An undo tablespace contains one or more undo segments. The creation of any other types of segment (for example, tables, indexes) in undo tablespaces is not allowed.
In the automatic mode, each Oracle instance is assigned one and only one undo tablespace. Each undo tablespace is composed of a set of undo files. Undo blocks are grouped in extents. At any point in time, an extent is either allocated to (and used by) a transaction table, or is free.
Blocks in undo tablespaces are grouped into the following categories:
Segments containing all undo data when the database is run in automatic undo management mode. Undo segments are internally similar to the rollback segments that are used to store undo data in manual undo management mode.
Like a rollback segment, an automatic undo segment must have a transaction table and two or more undo extents. Oracle stores the status of transactions in the transaction tables. Automatic undo segments are internally managed and cannot be manipulated by users through an external interface.
Backups made using a non-RMAN method, for example, using an operating system utility. For example, you can make a user-managed backup by running the
cp command on UNIX or the
copy command on Windows. A user-managed backup is also called an operating system backup.
A backup and recovery strategy for an Oracle database that does not use RMAN. This term is equivalent to operating system backup and recovery. You can back up and restore database files using operating system utilities (for example, the
cp command in UNIX), and recover using the SQL*Plus
RECOVER statement or the SQL
A test that checks whether a backup set or copy can be restored. RMAN scans all of the copies or backup pieces in the specified backup sets and looks at the checksum to verify that the contents can be successfully restored.
BACKUPSET command when you suspect that one or more copies or backup pieces in a backup set are missing or have been damaged. Note that
BACKUPSET actually test whether the files can be restored, whereas
CROSSCHECK merely examine the file headers.
A backup of the control file and all datafiles that belong to a database.
See Also: backup
Errors that occur when Oracle is unable to write data into a datafile, control file, or online redo log.
The Oracle protocol of recording changes to Oracle data blocks and undo segments in the online redo log before applying the changes to the actual data blocks. In other words, before database writer can write out a cache buffer containing a modified data block, log writer must write out the redo log buffer containing redo records describing both the changes to the data block as well as the undo required to roll back those changes if the transaction does not commit. This protocol ensures that changes can be recovered even if an instance failure occurs before the change can be written to the datafiles.