| Lesson 3 | Oracle RMAN Components and Architecture |
| Objective | Review the required and optional components of an Oracle AI Database 26ai RMAN environment. |
An Oracle Recovery Manager environment can be simple or extensive. At its minimum, it has only two required components: the RMAN client and the target database. The client accepts commands and coordinates the operation. Database server sessions associated with the target perform the physical backup, restore, recovery, validation, and maintenance work.
Other components are introduced when the recovery strategy needs them. Disk storage or a Fast Recovery Area can hold disk backups. An SBT library or media manager can connect RMAN to tape, Recovery Appliance, or a configured external destination. A recovery catalog can retain synchronized metadata and stored scripts. An auxiliary database or physical standby can participate in specialized operations. These components are optional or topology-dependent; they are not a fixed five-part RMAN installation.
| Component | Required? | Role |
|---|---|---|
| RMAN client | Yes | Interprets commands, establishes connections, directs server sessions, and reports results. |
| Target CDB or database | Yes | Contains the physical files and the authoritative RMAN repository in its control file. |
| Channel/server session | Required for I/O | Transfers one stream of data to or from a DISK or SBT device. |
| Disk or Fast Recovery Area | Optional destination | Stores disk backups and other recovery files visible to the target instance. |
| SBT media manager or native library | Optional destination path | Connects an SBT channel with tape, Recovery Appliance, or another configured destination. |
| Recovery catalog | Usually optional | Stores synchronized RMAN metadata, longer history, and stored scripts in an Oracle schema. |
| Auxiliary or standby database | Operation-dependent | Supports duplication, specialized recovery, or Data Guard protection workflows. |
The rman executable is the command-line client. It contains the RMAN command interpreter and accepts commands interactively or from a
client-side command file. It can run on the database host or another host that can reach the required databases through Oracle Net. The client establishes
target, catalog, or auxiliary connections as an operation requires.
The client does not read database blocks and write backup pieces by itself. Instead, it directs database server sessions to execute the requested work. This separation is important when diagnosing performance or connection problems: the client controls the operation, while the target or auxiliary instance supplies the sessions that perform physical I/O.
Oracle Enterprise Manager Cloud Control can provide a browser-based interface for database backup and recovery through RMAN. It is an optional management layer, not a required RMAN component. A managed OCI or Autonomous service may expose a different console, API, or responsibility boundary; direct customer-managed RMAN access must be verified in that service's documentation.
The target database contains the files RMAN protects or restores. In Oracle AI Database 26ai, the normal architecture is a container database (CDB) with a root and one or more pluggable databases (PDBs). RMAN operations can address the whole CDB or supported PDB, tablespace, data-file, control-file, SPFILE, and archived-redo scopes under the applicable connection and database-state rules.
The target control file always contains RMAN repository metadata. This metadata describes backups, image copies, archived redo logs, database incarnations, persistent configuration, and related activity. Oracle identifies the control-file records as the authoritative backup record for the target. Protecting the control file and enabling an appropriate control-file autobackup are therefore important parts of the recovery design.
The repository is metadata, not the backup data itself. Backup pieces and image copies reside on their configured devices. Likewise, a recovery catalog can supplement repository history, but it neither replaces the target control file nor becomes the destination for physical backup pieces.
An RMAN channel represents one stream of data to a device and corresponds to one database server session. During backup, a channel reads database blocks, processes them according to the backup command, and writes the output. During restore, it reads backup media and writes the restored file. The RMAN client directs these sessions but does not perform their block transfers.
“Server session” is more accurate than treating every connection as a separate operating-system process. On Linux, a server session commonly corresponds to a server process. On Windows, it can be a thread inside the Oracle database service. The logical RMAN architecture is the same even though the operating- system implementation differs.
Channels can be allocated automatically from persistent configuration or manually for a particular operation. RMAN is preconfigured with one automatic
DISK channel. Multiple channels can enable parallel work when the edition, operation, file layout, storage, network, CPU, and media resources
support it. More channels do not guarantee faster backups; excessive parallelism can move the bottleneck elsewhere.
RMAN> SHOW ALL;
RMAN> CONFIGURE DEFAULT DEVICE TYPE TO DISK;
RMAN> CONFIGURE CONTROLFILE AUTOBACKUP ON;
These settings persist in the RMAN repository for the connected target database. They establish defaults, not a complete recovery strategy. The administrator must still verify the destination, capacity, retention, credentials, failure domain, monitoring, and actual recovery behavior.
Oracle RMAN supports two device types: DISK and SBT. A DISK channel creates or reads backups in the filesystem namespace available
to the target instance. The destination can be a conventional filesystem, an ASM disk group, the Fast Recovery Area, or another supported server-visible
location. RMAN does not call a media manager for a DISK backup.
SBT means system backup to tape, but the interface is not limited to a physical tape drive. An SBT channel calls compatible media-management
software or an Oracle-provided native library. That external component manages its devices, storage objects, authentication, retention, and resource
allocation according to its product or service.
The architectural paths are therefore different:
An SBT channel should not be described as a “tape process,” and a cloud bucket should not be presented as a direct generic RMAN destination without its documented SBT library or service integration.
Consider what happens when the administrator submits a backup command. The RMAN client parses the command, checks its syntax, and determines which target connection and repository information apply. Persistent configuration can supply choices such as the default device type, automatic-channel settings, control-file autobackup behavior, and backup format. Those settings are read as instructions; they are not a separate backup engine.
RMAN then allocates the required target-database channels. Each channel is a server session that reads the selected database files. A DISK channel writes its output into the target instance's accessible disk namespace. An SBT channel passes requests through the configured SBT library, after which the media manager or integrated service controls the external destination. If several channels are allocated, the client coordinates multiple data streams while each channel maintains its own server session and device relationship.
As work completes, RMAN records the resulting backup metadata in the target control file. The records identify the created backup sets, pieces, or copies and associate them with the target database and its recovery history. When a recovery catalog is in use, RMAN synchronizes appropriate control-file metadata into the catalog schema. The catalog does not sit in the block-transfer path and does not receive a copy of the backup data.
Restore follows the same component boundaries in the opposite data direction. RMAN consults repository metadata, allocates channels capable of reaching the required device, reads the selected backup media, and writes restored files through target or auxiliary server sessions. Recovery can then apply the required incremental changes and redo. This end-to-end view is more accurate than treating “backup” or “recovery” as one process attached directly to every possible destination.
The Fast Recovery Area (FRA) is an optional disk location managed by Oracle Database and RMAN for recovery-related files. It can hold RMAN backups, archived redo logs, flashback logs, and copies of control files or online redo logs. When the FRA is configured and a disk-backup format is not supplied, RMAN can place backups there using Oracle Managed Files naming.
The FRA is a destination and space-management facility, not a separate repository. Its presence does not automatically create an offsite or independent copy. Backup files, archived logs, and flashback logs can compete for its configured capacity, so administrators must monitor usage and understand the deletion rules. A recovery plan should also address failures that affect both the production database and locally attached recovery storage.
A recovery catalog has two related components. The recovery catalog database is an Oracle database that hosts the catalog. The recovery catalog schema is the user that owns the RMAN-maintained metadata tables. One base catalog can contain metadata for multiple registered target databases and can support restricted virtual private catalog access.
RMAN synchronizes target control-file metadata into the catalog. The target database never accesses the catalog directly; the RMAN client maintains the
catalog connection and performs the propagation. The catalog can preserve a longer history than the reusable control-file records, support stored scripts,
and provide centralized RC_* reporting views.
The catalog is optional for many RMAN environments, but not for every topology. Oracle requires it when RMAN is used to back up a physical standby database. Starting with Oracle AI Database 26ai, RMAN automatically disconnects from the recovery catalog before performing a target-database backup operation, freeing the catalog connection while the backup work proceeds. This reinforces that the catalog supplies metadata rather than backup I/O.
A recovery catalog needs its own protection and availability plan. Losing the catalog does not erase existing backup pieces, but it can remove centralized history and stored scripts and can complicate operations that depend on catalog information.
An auxiliary database or instance is an operation-dependent component. RMAN uses an auxiliary connection for tasks such as
DUPLICATE, tablespace point-in-time recovery, table recovery, and certain transport or recovery workflows. The auxiliary environment may be
temporary, or it may become a persistent duplicate database.
A physical standby database is a Data Guard database updated by applying redo from its primary. RMAN can create, back up, restore, or recover a physical standby. Backups made at a physical standby can be usable at the primary or another physical standby associated with the same production database, subject to accessibility and repository rules.
A standby is not simply another backup destination like disk or tape. Data Guard provides standby-based availability and disaster recovery, while RMAN provides physical backup, restore, and recovery operations. The technologies complement one another but have different responsibilities.
An active RMAN session can contain more connections than the main architecture diagram needs to display. Depending on the operation, RMAN can maintain an initial target connection, one target connection for each allocated channel, a catalog connection, and an auxiliary connection. It also uses a polling connection to monitor command execution on the allocated channels.
By default, RMAN creates one polling connection. Additional polling connections can exist when configured or allocated channels use distinct connect strings. The polling connection monitors work; it does not read database blocks or write backup data. It belongs in troubleshooting and session-monitoring discussions rather than as a peer of disk, tape, or the catalog in an introductory architecture diagram.
Media-management software is needed only when the chosen destination uses SBT. A media manager can control tape drives, tape libraries, or another supported external system. It may maintain a vendor-specific media catalog. That catalog is distinct from the RMAN recovery catalog: the former tracks external media, while the latter stores Oracle RMAN metadata about registered target databases.
Oracle AI Database 26ai includes native SBT support for documented Oracle appliance and cloud-protection destinations. Native libraries simplify library deployment, but administrators must still configure credentials, network access, destination properties, retention, monitoring, and recovery procedures. Certified third-party SBT media managers remain possible for supported releases and platforms.
Oracle Secure Backup can continue to supply an SBT tape interface during its supported lifecycle. However, Oracle announced the desupport of Oracle Secure Backup 19.1 on May 1, 2026, with Premier Support ending September 30, 2027. New designs should check its current status and evaluate a supported media manager or Oracle cloud-protection option instead of treating OSB as the permanent default.
For applicable OCI protection requirements, Oracle recommends Zero Data Loss Cloud Protect using Autonomous Recovery Service. Recovery Appliance and Oracle Database Cloud Backup for Amazon S3 provide other documented paths. These products remain destination-specific integrations; they do not change the basic RMAN client, target, channel, and repository architecture.
Target and auxiliary connections require appropriate administrative privileges. For routine backup administration, Oracle recommends a designated account
granted SYSBACKUP rather than routine use of SYS. Starting with Oracle AI Database 26ai, dictionary protection is enabled for
SYSBACKUP.
Passwords and encryption secrets should not appear in command lines, command files, stored scripts, screenshots, or lesson examples. Recovery planning must separately protect the password file, SPFILE, encryption wallet or keystore, recovery-catalog credentials, media-manager configuration, and service credentials. A usable backup can still be unrecoverable if the required keys or external configuration are lost.
Data Recovery Advisor is not part of the current component model. Oracle AI Database 26ai desupports Data Recovery Advisor and the RMAN commands
LIST FAILURE, ADVISE FAILURE, REPAIR FAILURE, and CHANGE FAILURE. Oracle states that there is no
replacement feature, so neither Enterprise Manager nor ordinary validation should be described as its official successor.
| Decision | Question to resolve |
|---|---|
| Target scope | Does the strategy protect the complete CDB, selected PDBs, or another documented scope? |
| Device path | Will channels use DISK, SBT, or both, and can every recovery host reach the required media? |
| Repository | Is control-file history sufficient, or does the topology require or benefit from a recovery catalog? |
| External components | Which media manager, native SBT library, auxiliary instance, standby, or service is actually required? |
| Security | Are SYSBACKUP, credentials, wallets, keys, and configuration protected and recoverable? |
| Operations | Are capacity, job results, repository state, retention, and destination availability monitored? |
| Recovery evidence | Has the complete restore-and-recovery path been tested against the required recovery objectives? |
A clear RMAN design lets an administrator identify who issues commands, where work executes, where metadata resides, how each channel reaches its device, and which optional systems participate. The next lesson examines the preparation and configuration decisions that should be completed before routine RMAN operations begin.