Configuring Recovery Manager and Archiving

You can configure Oracle Recovery Manager (Oracle RMAN) to support your Oracle RAC environment.

This chapter provides procedures for using for archiving in Oracle RAC environments and discusses online redo log and archived redo log considerations.

Overview of Configuring RMAN for Oracle RAC

Oracle Recovery Manager (RMAN) is included with Oracle AI Database and supports backup, restoration, and recovery of data files, control files, SPFILEs, and archived redo logs in Oracle RAC. It can be used from the command line or through Oracle Enterprise Manager.

Archiving Mode in Oracle RAC

An Oracle RAC database must use ARCHIVELOG mode to archive redo logs. Archiving mode is configured at the database level, so all instances either archive logs or none do; it can be changed with ALTER DATABASE while the database is mounted and not open.

Configuring the RMAN Snapshot Control File Location

RMAN uses a snapshot control file to obtain a consistent control-file copy for recovery catalog synchronization and control-file backups. In Oracle RAC, its location must be on shared storage accessible to every backup node; SHOW SNAPSHOT CONTROLFILE NAME displays the location, while CONFIGURE SNAPSHOT CONTROLFILE NAME changes it persistently.

Configuring RMAN to Automatically Backup the Control File and SPFILE

Enabling CONFIGURE CONTROLFILE AUTOBACKUP causes RMAN to back up the control file and SPFILE after BACKUP or COPY operations. Autobackups support disaster recovery without a recovery catalog, but their storage path must be identically accessible on all participating nodes, and the first allocated RMAN channel performs the autobackup.

Crosschecking on Multiple Oracle RAC Nodes

For reliable RMAN crosschecks and restores, every Oracle RAC node should be able to access all backups. If this is not possible, configure channels with CONNECT for the nodes that can access particular backups; otherwise inaccessible backups may be incorrectly marked EXPIRED during crosschecking.

Configuring Channels for RMAN in Oracle RAC

RMAN channels in Oracle RAC can be configured for automatic load balancing or assigned to specific instances. Automatic balancing is enabled by configuring device-type parallelism with CONFIGURE DEVICE TYPE, specifying the number of channels RMAN should use for backup or restore operations.

Managing Archived Redo Logs Using RMAN in Oracle RAC

Oracle records archived redo log filenames in the target control file and, when applicable, the recovery catalog. Archived logs must use naming and storage arrangements that allow required nodes to read them; although multiple nodes can restore logs in parallel, the node applying recovery must access every required log, with RECOVER PARALLEL available to adjust recovery threads.

Archived Redo Log File Conventions in Oracle RAC

LOG_ARCHIVE_FORMAT must uniquely identify archived redo logs in Oracle RAC using thread, sequence, and resetlogs variables such as %t, %s, and %r, with uppercase variants providing zero padding. Including the resetlogs identifier prevents files from different database incarnations from being overwritten, and these format variables are mandatory when COMPATIBLE is 10.0 or higher.

RMAN Archiving Configuration Scenarios

Oracle RAC archiving can use shared cluster file systems or node-local file systems with remote access. Shared storage, preferably Oracle ASM with a separate recovery-area disk group, lets every node read all logs directly and simplifies backups; local archiving requires unique per-node destinations plus NFS or mapped-drive access during recovery. In either design, archived logs must be complete and readable by the recovery node, and the recovery instance must use the same LOG_ARCHIVE_FORMAT as the archiving instances.

Monitoring the Archiver Processes

After configuring RMAN and archiving, use GV$ARCHIVE_PROCESSES to view archiver status across all Oracle RAC instances or V$ARCHIVE_PROCESSES to view the connected instance. Do not terminate an archiver process with the kill command because doing so causes the database instance to fail.