Join our Newsletter — 33% off our NHI Course
Home› Guides› Oracle Critical Patch Update Guide: Planning Quarterly CPUs,…
Guide Vulnerability Management

Oracle Critical Patch Update Guide: Planning Quarterly CPUs, Monthly CSPUs and Alerts

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 17 min read
On this page

Oracle estates are patched on Oracle's own calendar, and teams that treat the quarterly Critical Patch Update as a once a quarter scramble end up running database, middleware and Java homes that are one or more cycles behind. Oracle publishes Critical Patch Updates (CPUs) on the third Tuesday of January, April, July and October, and since May 2026 it has added monthly Critical Security Patch Updates (CSPUs) in the months between. This guide explains how the programme works, how to read a CPU advisory, how to turn it into a repeatable plan for Oracle Database, Fusion Middleware and WebLogic, and Java SE, and how to measure whether the plan is holding. It is written for vulnerability managers, DBAs, middleware administrators and change managers who own Oracle on-premises software.

This is an independent guide from NHI Mgmt Group, based on Oracle's public documentation as of November 2026. It is not sponsored or reviewed by Oracle. Product features change, so confirm settings against Oracle's current documentation.

Key takeaways

  • Put the dates in the change calendar a year ahead. CPUs land on the third Tuesday of January, April, July and October at around 1 p.m. Pacific Time, CSPUs on the third Tuesday of the other eight months, and the pre-release announcement on the Thursday before each.
  • CPUs are cumulative, CSPUs are not. Oracle's FAQ states that CSPUs are not cumulative, while each quarterly CPU includes the fixes from prior CSPUs, so the quarterly CPU remains the baseline you plan around.
  • Only supported releases get fixes. CPU patches are produced for releases in Premier or Extended Support, and for most products only the latest versions within each release, so lifecycle position decides whether you can patch at all.
  • For Oracle Database, stay on the latest Release Update. Oracle's patch maintenance guidance recommends applying the latest quarterly RU, out of place, with Monthly Recommended Patches or CSPUs as the option when you run one RU behind.
  • Security Alerts break the cycle. Oracle issues a Security Alert for fixes too critical to wait for the next CPU or CSPU; your process needs an emergency path that does not wait for the quarterly window.

How the Oracle programme works

Oracle's vulnerability fixing policy describes the CPU as the primary mechanism for backporting security fixes to all Oracle on-premises products. Oracle generally announces a fix only when it is available for all affected supported version and platform combinations, with exceptions for the On-Request programme and short patch delays of up to two weeks after the announcement. Oracle states that it gives the same information to all customers, with no advance notice to individual customers, and that it does not distribute exploit or proof of concept code. The table below sets out each release vehicle as Oracle's public pages describe it.

VehicleCadenceCumulative?What it is for
Critical Patch Update (CPU)Quarterly, third Tuesday of January, April, July, OctoberYes, as far as possible; includes all prior CPU and CSPU fixesThe planned baseline for every Oracle on-premises product
Critical Security Patch Update (CSPU)Monthly, third Tuesday of February, March, May, June, August, September, November, DecemberNoTargeted, high-priority fixes in a smaller format between CPUs (first released 28 May 2026)
Security AlertOut of band, as neededIncluded in the next CPU unless notedFixes judged too critical to wait for the next CPU or CSPU; customers are notified by email and through My Oracle Support
Pre-release announcementThursday before each CPU, and before each CSPU from June 2026Not applicableAffected products, fix counts per product family and the highest CVSS base score per family, for planning only
Solaris Third Party BulletinSame day as CPUs, updated once between quartersNot applicableThird-party software shipped with Oracle Solaris
Oracle Linux security advisories (ELSA)ContinuousNot applicableOracle Linux and Oracle VM Server for x86 receive updates through these rather than CPUs

As published on Oracle's Security Alerts page in November 2026, the next dates are a CSPU on 17 November 2026 and 15 December 2026, a CPU on 19 January 2027, CSPUs on 16 February and 16 March 2027, and CPUs on 20 April, 20 July and 19 October 2027. Oracle's fixing policy page and its own month table do not fully agree on CSPU wording, so take dates from the Security Alerts page each year rather than from memory.

Reading a CPU advisory

Each advisory opens with an Affected Products and Patch Information table. One column lists products and versions, linked to per-family appendices; the other links each family to its Patch Availability Document on My Oracle Support, where the exact patch numbers for each release and platform live. Access to the patches themselves requires an active support contract. The parts that drive planning are public:

  • Risk matrices: one per product family (Database Server, Fusion Middleware, Java SE, MySQL, E-Business Suite and others). Columns are CVE ID, Component, Package and/or Privilege Required, Protocol, Remote Exploit without Auth., CVSS 3.1 Base Score, the CVSS base metrics, Supported Versions Affected and Notes. Matrices are ordered by base score, most severe first.
  • Remote Exploit without Auth.: the single most useful triage column. A "Yes" on a component you expose to a network is the first thing to schedule.
  • Protocol: Oracle states that the protocol shown implies all its secure variants are affected too, so a fix listed against HTTP also applies to HTTPS.
  • Supported Versions Affected: match this against your inventory, down to the RU or patch set level, not just the major version.
  • Additional patches and VEX: for bundled third-party components Oracle lists VEX justifications such as component_not_present or vulnerable_code_not_in_execute_path. Oracle has published VEX justifications since the July 2023 CPU, and the advisories are also available in CSAF JSON, including by RSS, which lets a vulnerability platform ingest them without screen scraping.
  • Workarounds and skipped patches: the advisory states that workarounds and mitigations are not a long-term solution because they do not correct the underlying problem. Because CPUs are cumulative, customers who skipped earlier cycles should review the earlier advisories for products whose fixes are not cumulative.

Oracle uses CVSS 3.1 in current advisories, adopted in July 2020. For prioritisation beyond the base score, combine it with exposure and with exploitation likelihood; see the EPSS Guide and the CIS Control 7 Continuous Vulnerability Management Guide.

A quarterly CPU routine

The pattern that works is the same each quarter, so it can be written once and run by rota. Compared with a monthly patch cycle, the difference is fewer releases, larger bundles and longer test cycles for stateful database and middleware tiers.

  1. T minus 5 days (Thursday pre-release): read the pre-release announcement, note which product families are affected and their highest CVSS base score, and confirm test and production change windows for the coming weeks.
  2. Release day (Tuesday, around 1 p.m. Pacific): read the advisory, filter the risk matrices to products in your inventory, and flag every "Remote Exploit without Auth. = Yes" row on an internet-facing or widely reachable component.
  3. Day 1 to 2: pull the Patch Availability Documents for each family you run, record the exact patch numbers per Oracle home and platform, and check for conflicts with existing interim patches. Oracle's FAQ directs conflict questions to Oracle Support.
  4. Day 2 to 10: build patched Oracle homes out of place, apply to development and test, run regression and performance tests, and confirm datapatch or post-patch SQL completed.
  5. Day 10 to 30: roll to production in rings: standby or non-critical first, then primary, using rolling methods where the platform supports them.
  6. Day 30 onwards: verify through inventory (opatch lsinventory, DBA_REGISTRY_SQLPATCH) that every home is on the new level, raise exceptions for any that are not, and retire the old homes once rollback is no longer needed.

The day ranges are a starting point, not Oracle guidance; set them in your own policy against risk tolerance. Oracle says installing CPUs is not mandatory and recommends that customers test patches in their own test environments, while also recommending that CPU, CSPU and Security Alert fixes are applied without delay.

Oracle Database: RUs, MRPs and CSPUs

Oracle Database proactive patches arrive as Release Updates (RUs) on the same quarterly dates as the CPU. RUs are cumulative, highly tested bundles; the release number shows the RU level, so 19.7.0.0.0 is the seventh quarterly RU for 19c. Release Update Revisions were discontinued after January 2023, and Monthly Recommended Patches (MRPs) took their role from RU 19.17 onwards. MRPs apply to 19c on Linux x86-64, are released monthly for up to six months after each RU, do not change the release number, and may include CPU alert and high CVSS fixes but not proactive functional fixes.

Policy option (from Oracle's guidance)What you runWhen it fits
Quarterly, NLatest RU each quarter, about three months per RUOracle's stated security best practice and the default for single-instance databases
Quarterly, N-1Prior RU, with continuous monitoring and accelerated patching for material fixesEstates that want one cycle of field exposure before adopting new RU content
Monthly, N-1Latest MRP for the prior RU every monthLinux x86-64 19c estates that can absorb monthly change
Monthly Long Term RU, N-1A Long Term RU for 12 months, plus its latest MRP, moved during an annual change windowHighly change-averse systems; pair with strong compensating controls

Oracle's guidance is explicit that if N-1 is used to reduce early-adoption risk it should be combined with MRPs or CSPUs, and that higher CVSS levels, especially Critical (9.0 to 10.0) and High (7.0 to 10.0), may require accelerated adoption of a newer RU, MRP, CSPU or interim fix. Oracle grades software health as GREEN (no action), YELLOW (behind by one cycle or missing important fixes) and RED (behind by more than one cycle or missing critical fixes), and recommends checking weekly, more often than you apply.

  • Method: Oracle recommends that all patch operations are performed out of place, with DBCA for single-instance databases and Fleet Patching and Provisioning (FPP) for RAC, Data Guard and larger fleets, using gold images. Oracle Update Advisor supplies policy-aware recommendations; dbca -managePatches -checkPatchStatus reports status against your policy.
  • Zero downtime: both RUs and MRPs support RAC rolling application.
  • Before the window: stop Data Pump jobs and other data movement, and shut down GoldenGate processes before patching, as Oracle's patch maintenance guide requires.
  • Verify: opatch lsinventory for binaries and DBA_REGISTRY_SQLPATCH for SQL patches in each database.
  • Combo and OJVM: Oracle also ships OJVM RU plus Database RU combo patches, quarterly full stack downloads for Exadata and time zone patches every six months; include them in the inventory so they are not forgotten.

For database hardening around the patch cycle, see the Oracle Database Security Guide.

Fusion Middleware and WebLogic

Middleware is where CPU slippage tends to concentrate, because many Oracle homes sit under packaged applications with their own certification lag. OPatch is the utility for applying Fusion Middleware patches. Its documentation lists interim patches, bundle patches, Security Patch Updates (formerly known as CPU) and Patch Set Updates, which are used for WebLogic Server only.

  • Back up the Oracle home: Oracle's OPatch documentation highly recommends a backup of the Oracle home before any patch operation.
  • Use the OPatch in the Oracle home: Fusion Middleware customers must use the OPatch version shipped in the Oracle home and move to another version only when told to by support. Check it with opatch version.
  • Pre-check: run opatch checkApplicable, or opatch apply with -report to see the actions without applying them.
  • Roll back cleanly: opatch rollback or nrollback removes patches; logs are under ORACLE_HOME/cfgtoollogs/opatch.
  • Zero Downtime Patching: WebLogic Server 14.1.2 ZDT workflows move a domain from an existing Oracle home to a patched one, update the Java version and perform rolling restarts without loss of service. This is the out-of-place pattern for clustered domains.
  • Java under middleware: the JDK beneath WebLogic is a separate patch target. A patched WebLogic home on an unpatched JDK still carries Java SE fixes you have not applied.

See the Oracle WebLogic Server Hardening Guide for the configuration controls that reduce exposure between cycles.

Java SE in the CPU cycle

Java SE appears in each CPU advisory with its own risk matrix and patch availability document. The planning issue is less the fixes themselves than the version and licence position of every JDK and JRE in the estate. Oracle's Java SE Support Roadmap, as checked in November 2026, lists 8, 11, 17, 21 and 25 as long-term support releases, with non-LTS feature releases each March and September superseded as soon as the next one ships. Oracle states that updates to Oracle JDK 17 and 21 after set dates fall under the Java SE OTN licence, while JDK 25 and later are available under the No-Fee Terms and Conditions licence until the dates the roadmap gives.

  • Inventory every runtime: application servers, batch hosts, desktop clients, build agents and runtimes bundled inside third-party products.
  • Retire non-LTS releases: a non-LTS release receives no further updates once superseded, so it falls behind at the next CPU.
  • Check entitlement before release day: know which runtimes you can update and under which licence, so the CPU is not delayed by a procurement question.

See Java End of Life Versions for lifecycle dates by version.

Decision guidance

SituationRecommended approachTrade-off
Internet-facing WebLogic, ORDS or E-Business Suite with "Remote Exploit without Auth. = Yes" rowsEmergency or accelerated change within days, CSPU or interim fix if the CPU is weeks awayLess regression testing; mitigate with out-of-place homes and fast rollback
Core OLTP database behind several network layersLatest RU quarterly, RAC rolling or standby firstQuarterly change load on DBAs
Change-averse system that cannot move every quarterN-1 RU plus MRPs or CSPUs, documented exception with compensating controlsRuns behind Oracle's best practice; needs active monitoring for alerts
Packaged application certified only on older middlewareFollow the application vendor's certified CPU path; escalate if it lags more than one cycleDependency on a third party's timetable
Release in Sustaining Support onlyPlan the upgrade; no new CPU fixes are produced for releases outside Premier or Extended SupportUpgrade cost against unpatchable risk

Small organisations with a handful of Oracle homes can run the routine manually with a spreadsheet inventory and the quarterly dates in a shared calendar. Large estates need automated health checks, FPP or equivalent fleet tooling, and a CSAF feed into the vulnerability platform so advisories map to assets without manual reading.

Where identity fits

Oracle patching runs with powerful accounts: the OS owner of each Oracle home, SYSDBA connections for datapatch, WebLogic administrators for ZDT workflows, and the service accounts behind fleet tooling. These are high-value non-human identities. Vault their credentials, scope them to the hosts they patch, and log their use during each window. See the Oracle Database Security Guide for privileged database account controls.

Detection and measurement

  • CPU currency: percentage of Oracle homes on the current quarterly level (latest RU or the current CPU patch set) 30 and 60 days after release.
  • Cycles behind: count of homes YELLOW (one cycle behind) and RED (more than one), using Oracle's own health grading.
  • Exposure time on critical rows: median days from advisory to patch for components with "Remote Exploit without Auth. = Yes" and a CVSS base score of 9.0 or more.
  • Unsupported footprint: number of homes and JDKs on releases outside Premier or Extended Support, trending to zero.
  • Exception age: open CPU exceptions older than one quarter, each with an owner and compensating control.
  • Rollback rate: production patch rollbacks per cycle, a signal of test coverage.

Feed opatch lsinventory output, DBA_REGISTRY_SQLPATCH and JDK version scans into the CMDB weekly, and alert on any home whose level drops behind the policy target.

Framework mapping

FrameworkReferenceHow the CPU routine meets it
NIST SP 800-53 Rev 5SI-2 Flaw Remediation; SI-2(2) Automated Flaw Remediation Status; SI-2(3) Time to Remediate FlawsQuarterly RU and CPU application, automated health checks, defined remediation times
NIST SP 800-53 Rev 5SI-5 Security Alerts, Advisories, and Directives; RA-5 Vulnerability Monitoring and ScanningSubscribing to Oracle advisories, alerts and CSAF; inventory-based matching
NIST SP 800-53 Rev 5CM-3 Configuration Change Control; CM-8 System Component InventoryPlanned CPU windows, emergency path for Security Alerts, Oracle home and JDK inventory
NIST CSF 2.0ID.RA-01; ID.AM-02; PR.PS-02Vulnerabilities identified per asset; software inventory; software maintained and replaced according to risk
CIS Controls v8.17.3 and 7.4 automated operating system and application patch managementFleet tooling and gold images for Oracle homes
ISO/IEC 27001:2022Annex A 8.8 Management of technical vulnerabilities; 8.32 Change managementTimely evaluation and patching under change control

For a programme-level view of patch planning, see the NIST SP 800-40 Rev 4 Patch Management Guide.

Common failure patterns

  • Patching the database, forgetting the stack: RU applied to the database home while the Grid Infrastructure home, OJVM component, client homes and the JDK under middleware stay on old levels.
  • Treating CSPUs as cumulative: skipping a CSPU and assuming the next one carries it. CSPUs are not cumulative; only the quarterly CPU rolls them up.
  • Binary patched, SQL not: opatch reports success but datapatch was never run or failed in one pluggable database, so DBA_REGISTRY_SQLPATCH shows the old level.
  • Running out of support: releases outside Premier or Extended Support do not receive new CPU fixes, and Oracle warns that attackers often reverse engineer fixes shortly after release, so these systems carry known, unfixable issues.
  • Waiting a full quarter on a Security Alert: holding an out-of-band fix for the next scheduled window instead of using the emergency change path.
  • In-place patching with no rollback home: patching the live Oracle home without a backup or a parallel home, so a failed patch becomes an outage.

Practitioner checklist

  • Add the next four CPU dates, eight CSPU dates and their Thursday pre-release dates to the change calendar.
  • Subscribe owners to Oracle Security Alert emails and ingest the CSAF advisories into the vulnerability platform.
  • Keep an inventory of every Oracle home, its RU or patch level, and every JDK and JRE, refreshed weekly.
  • Write a database maintenance policy choosing quarterly N, N-1 with MRPs or CSPUs, or Long Term RU, with apply frequency and notification level.
  • Move to out-of-place patching with gold images: DBCA or FPP for databases, ZDT workflows for WebLogic clusters.
  • Define an emergency change path for Security Alerts and critical CSPUs that does not wait for the quarterly window.
  • Verify every cycle with opatch lsinventory, DBA_REGISTRY_SQLPATCH and JDK version scans, and close exceptions with owners and dates.
  • Track and plan upgrades for any release leaving Premier or Extended Support.

Standards and references

Related NHI Mgmt Group resources: Oracle Database Security Guide · Oracle WebLogic Server Hardening Guide · Java End of Life Versions · NIST SP 800-40 Rev 4 Patch Management Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.