Apply Software Updates

Well-planned, tested, and properly executed software updates are crucial to maximizing uptime and minimizing business disruption. Wherever possible, Oracle recommends out-of-place, gold image-based patching—especially for clustered systems—to ensure safe rollbacks and rapid issue resolution.

Key Practices

• Before updating software run Exachk/Orachk and correct configuration that does not follow recommended best practices.

• Follow client failover best practices to reduce application-level impact during rolling updates that disrupt database instance connectivity. See Continuous Availability for Applications for details.

• Test the update and rollback process in a non-production environment with production-like workload and configuration.

• Use Data Guard Standby-First Patch Apply approach to minimize risk and primary database impact when updating mission critical systems. See Oracle Patch Assurance - Data Guard Standby-First Patch Apply (KB137118) for details.

• Take system and database backups immediately before software changes.

• Always validate, backup, and have a proven rollback plan before making software changes.

• Use repeatable, automated, image-based workflows provided by Oracle software maintenance tools to ensure consistency and reduce manual errors.

• Use DBCA/DBCACTL or AutoUpgrade for out-of-place, gold image-based Database updates on standalone/non-clustered systems.

• Use FPP or EM for coordinated, out-of-place, rolling, gold image-based Database and Grid Infrastructure updates on clustered environments such as RAC and Exadata.

• Apply online/in-place updates only for specific patch types or situations where minimal disruption is required. See MOS note RDBMS Online Patching Aka Hot Patching (KB140524) for details.

Software Update Method Selection: Benefits and Tradeoffs

Table 1 Database Software on Non-Clustered Systems (Standalone Hosts)

Method	Benefits/Tradeoffs
Out-of-place (gold image) with DBCA/DBCACTL or AutoUpgrade (Recommended)	Fast, simple, easy rollback Requires extra disk space $ dbca -moveDatabase -sourceDB -silent
Online hot patch with OPatch	No database downtime or database connection disruption Patching is done in-place in the existing database home Limited to supported patches (MRP/interim)

Table 2 Database Software on Clustered Systems (Oracle RAC/Exadata)

Method	Benefits/Tradeoffs
Out-of-place (gold image), RAC rolling with FPP or EM (Recommended)	Fast, simple, full cluster orchestration, easy rollback Requires extra disk space $ rhpctl move database -sourcewc -patchedwc
Online hot patch with FPP	No database downtime or database connection disruption Patching is done in-place in the existing database home Limited to supported patches (MRP/interim)
Non-rolling with FPP	All updates applied at once, reduced planned maintenance time Full database downtime

Table 3 Grid Infrastructure Software on Clustered Systems (Oracle RAC/Exadata)

Method	Benefits/Tradeoffs
Out-of-place (gold image), Oracle RAC rolling with FPP (Recommended)	Fast, simple, full cluster orchestration, easy rollback Requires extra disk space $ rhpctl move gihome -sourcewc -destwc
Zero-downtime out-of-place (gold image), Oracle RAC rolling with FPP	No database connection disruption Requires extra disk space $ rhpctl move gihome -sourcewc -destwc -tgip

Table 4 Exadata System Software

Method	Benefits/Tradeoffs
Rolling update with patchmgr or FPP (Recommended) Storage servers Database servers Network fabric switches	No database downtime, simple, full system orchestration $ patchmgr --dbnodes --repo --upgrade --target_version --rolling
Live update with patchmgr or FPP Database servers	No database connection disruption, supports partial updates to address security issues Outstanding work update items cannot be completed online and require a system reboot $ patchmgr --dbnodes --repo --upgrade --target_version --rolling --live-update-target full
Non-rolling with patchmgr or FPP Storage servers Database servers Network fabric switches	All updates applied at once, reduced planned maintenance time Full system downtime