Administering Database Instances and Cluster Databases

This chapter describes how to administer Oracle Real Application Clusters (Oracle RAC) databases and database instances.

Note:

A multitenant Container Database (CDB) is the only supported architecture in Oracle Database 21c and later releases.

Overview of Oracle RAC Database Administration

Oracle RAC administration uses the SYSRAC privilege and tools such as SRVCTL, Oracle Enterprise Manager, and SQL*Plus. Since Oracle Database 21c, administrator-managed and policy-managed capabilities have converged into a single CDB/PDB-oriented model, while policy-managed deployment is desupported in Oracle AI Database 26ai. The model supports dynamic PDB placement, scalable cluster resources, automatic redo and undo provisioning, and SCAN-based client connectivity.

Tools for Administering Oracle RAC

Oracle RAC is administered primarily with SRVCTL, Oracle Enterprise Manager, and SQL_Plus. SRVCTL centrally manages databases, instances, services, and Clusterware resources; Oracle Enterprise Manager provides graphical, cluster-wide and instance-specific monitoring and administration; and SQL_Plus performs operations against the current instance, requiring explicit instance targeting for remote or multi-instance changes. Use SYSASM rather than SYSRAC for ASM administration.

Starting and Stopping Instances and Oracle RAC Databases

Oracle RAC instances and databases can be started or stopped with Oracle Enterprise Manager, SRVCTL, or SQL_Plus, provided Oracle Grid Infrastructure is running on the target node. Startup state determines available operations, and Oracle does not support running multiple instances of the same database on one node. SRVCTL and Enterprise Manager can manage all instances together, while SQL_Plus is primarily instance-oriented.

Starting and Stopping PDBs in Oracle RAC

PDBs are Oracle Clusterware resources in Oracle Database 21c and later, so SRVCTL manages their startup, shutdown, placement, services, policies, cardinality, and status. Automatic PDBs start with their CDB, while manual policies preserve appropriate state across Clusterware restarts. Administrators can start or stop PDBs cluster-wide or on selected nodes, disable automatic restart, inspect configuration, and adjust cardinality and resource settings subject to creation-time restrictions.

Local Rolling Database Maintenance

Oracle AI Database 26ai supports optional local rolling, out-of-place maintenance for Oracle RAC and RAC One Node, allowing a second instance from a new Oracle home to run temporarily on the same node before the old instance stops. This reduces connection migration and brownout time while preserving PDB placement and services. The feature requires OMF, an SPFILE, sufficient CPU, memory, storage, redo, and undo capacity, and controlled SRVCTL configuration and instance transfer; failed transfers can be fully or partially reverted.

Pluggable Database Rank

The PDB -rank setting assigns relative workload importance to cardinality-based PDBs under the RANK management policy, with higher values receiving priority. Oracle Clusterware uses rank to determine startup order, instance shutdown decisions, and which PDBs remain running when resources are insufficient or nodes fail. Ranking is cluster-wide and is evaluated before CPU requirements during failover and resource allocation.

Pluggable Database Placement

Oracle RAC supports preferred and available PDB placement, which restricts a PDB to explicitly designated CDB instances or nodes, and floating placement, which allows a cardinality-defined PDB to run dynamically on eligible CDB instances. Cardinality controls the simultaneous node count, while -maxcpu and -mincpuunit influence resource evaluation. Placement can be configured when creating a PDB or modified later, with privileged users required for certain resource and ranking changes.

Example of Creating a Pluggable Database with Cardinality and Rank

The examples demonstrate creating a database with the RANK management policy, adding a PDB with a specified cardinality, inspecting its configuration, and modifying its cardinality, CPU limits, and rank using SRVCTL. A CDB using RANK defaults newly created PDBs to the RESTART policy, and PDB rank and resource controls help Oracle Clusterware prioritize workloads.

Reducing Downtime During Database and Instance Outages

Oracle RAC maintains availability during individual instance or server failures through automatic restart, recovery, subsystem management, service relocation, and client redirection. Planned work can be performed through rolling instance maintenance, while Clusterware service management, FAN, and Application Continuity reduce disruption by relocating services, notifying clients rapidly, draining sessions, balancing connections, and replaying recoverable requests.

Oracle RAC High Availability Best Practices

Oracle AI Database 26ai introduces two-stage rolling updates, separating patch deployment from later activation through ALTER SYSTEM ENABLE RAC TWO-STAGE ROLLING UPDATES ALL, with status visible in V$RAC_TWO_STAGE_ROLLING_UPDATES. Smooth reconfiguration reduces brownout during node changes by performing on-demand Cache Fusion reconfiguration, and ordered sequence optimizations improve RAC performance automatically. Sequence configuration should balance ordering, gaps, and performance, with cached NOORDER generally providing the best throughput.

Verifying That Instances are Running

Instance availability can be checked with SRVCTL, SQL_Plus, or Oracle Enterprise Manager._ srvctl status database reports running instances and nodes, while srvctl status pdb _checks PDB status. In SQL_Plus, administrators can connect to an instance-specific service and query V$ACTIVE_INSTANCES to view instance numbers, host names, and instance names.

Terminating Sessions On a Specific Cluster Instance

To terminate a RAC session, identify its SID, serial number, and instance ID in GV$SESSION, then issue ALTER SYSTEM KILL SESSION with the optional @instance_id clause. The operation rolls back active transactions and releases resources; it may wait for outstanding activity or mark the session for PMON termination, while IMMEDIATE forces prompt termination. Application Continuity can mask recoverable session cancellation from users.

Overview of Initialization Parameter Files in Oracle RAC

Oracle RAC instances should use one shared SPFILE, stored on ASM or a cluster file system, to maintain consistent and persistent configuration. Instance-specific values use SID.parameter entries, while wildcard values apply broadly; administrators should modify the binary SPFILE through Enterprise Manager or ALTER SYSTEM, not a text editor. The SPFILE location is managed through Clusterware, parameter files follow platform-specific search orders, and regular backups through RMAN or CREATE PFILE are recommended.

Initialization Parameter Use in Oracle RAC

Most initialization parameters use common defaults, but some may differ by instance and others must be identical across the cluster. RAC-specific settings include cluster mode, instance identity, redo threads, GCS processes, interconnects, shared server dispatchers, result cache behavior, services, and shared SPFILE usage. Critical parameters such as database identity, control files, block size, recovery destinations, instance type, undo management, and compatibility must match, while several operational parameters should also be aligned to simplify administration and avoid failover or recovery problems.

Quiescing Oracle RAC Databases

ALTER SYSTEM QUIESCE RESTRICTED quiesces the entire RAC database, not merely the issuing instance, and requires Database Resource Manager to have been active since startup on every instance. The command waits for non-DBA sessions to become inactive, prevents database opening and resource-plan changes during quiescence, and is reversed with ALTER SYSTEM UNQUIESCE. A quiesced database is unsuitable for cold backups because background processes may still update internal structures, although online backups remain possible.

Administering Multiple Cluster Interconnects on Linux and UNIX Platforms

On Linux and UNIX, CLUSTER_INTERCONNECTS can specify one or more alternative private-network addresses, distributing RAC traffic across listed interfaces when available. Oracle generally recommends Redundant Interconnect Usage instead, and the parameter should not contain HAIP addresses. When used, addresses must correspond to equivalent adapters across instances and be listed consistently; appropriate use cases include separating traffic for multiple databases or supporting unstable interface naming, while single-interconnect or adequately provisioned default configurations should leave it unset.

Customizing How Oracle Clusterware Manages Oracle RAC Databases

SRVCTL database policies control how aggressively Oracle Clusterware restarts RAC databases. AUTOMATIC restores the prior running state and restarts after failures, MANUAL avoids host-restart recovery while continuing failure monitoring, and NORESTART prevents automatic restarts even after failures. Administrators can view policies with srvctl config database -all, change them with srvctl modify database, or specify them when adding a database.

Advanced Oracle Enterprise Manager Administration

Oracle Enterprise Manager Cloud Control supports centralized installation, configuration, monitoring, and advanced administration of Oracle RAC databases, including node and instance discovery, provisioning, cluster scaling, job scheduling, alert management, and maintenance suspensions.

Using Oracle Enterprise Manager Cloud Control to Discover Nodes and Instances

Oracle Enterprise Manager Cloud Control can discover Oracle RAC database and instance targets automatically during agent installation or manually afterward through the Targets and Database pages. Administrators can add database, listener, and ASM targets, select hosts, choose discovery scope, and manually configure cluster databases or instances that remain undiscovered after reconfiguration.

Other Oracle Enterprise Manager Capabilities

Oracle Enterprise Manager provides RAC and Grid Infrastructure provisioning with reusable deployment profiles, dynamically updated prerequisites and tools through My Oracle Support, support for one-click cluster extension, certified cluster scale-down and deletion, single-instance database provisioning, and Grid Infrastructure provisioning for standalone servers.

Administering Jobs and Alerts in Oracle RAC

The Cluster Database Home page aggregates RAC instance status, alerts, and AWR statistics while allowing drill-down to individual instances. Through the Administration tab, administrators can manage RAC jobs and alerts, configure instance- or database-level thresholds, and use maintenance suspensions to prevent unnecessary monitoring noise during planned operations.

Administering Jobs in Oracle RAC

Oracle Enterprise Manager jobs can be created at the cluster database, cluster, or instance level and executed on an available or designated RAC host. Administrators can submit operating system and SQL scripts, schedule recurring tasks such as backups, and review job activity through Active, History, and Library categories.

Administering Alerts in Oracle RAC with Oracle Enterprise Manager

Oracle Enterprise Manager supports RAC-specific tests and distinguishes database-level alerts from instance-level alerts. Administrators can set thresholds for individual instance conditions, such as archive log issues, or configure database-wide conditions, such as tablespace usage, to avoid duplicate notifications across instances.

Using Defined Suspensions in Oracle Enterprise Manager

Defined suspensions temporarily disable monitoring for all RAC targets or selected cluster database instances during maintenance. This prevents planned operations from distorting monitoring data or generating unnecessary alerts.