<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ashoka Reddy Tatiparthi on Medium]]></title>
        <description><![CDATA[Stories by Ashoka Reddy Tatiparthi on Medium]]></description>
        <link>https://medium.com/@ashoka.dba?source=rss-05f8b1c04f1b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*YTQ6azmqJLonvhPD</url>
            <title>Stories by Ashoka Reddy Tatiparthi on Medium</title>
            <link>https://medium.com/@ashoka.dba?source=rss-05f8b1c04f1b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 06:32:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ashoka.dba/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Oracle 26 Ai Data Guard Fast role transitions]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-26-ai-data-guard-fast-role-transitions-30fbdd1f2f72?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/30fbdd1f2f72</guid>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Mon, 12 Oct 2026 12:55:58 GMT</pubDate>
            <atom:updated>2026-10-12T12:55:58.652Z</atom:updated>
            <content:encoded><![CDATA[<p>[oracle@ol9–26-dg1 ~]$ dgmgrl sys@cdb1 as sysdba<br>DGMGRL for Linux: Release 23.26.1.0.0 — Production on Mon Oct 12 12:27:25 2026<br>Version 23.26.1.0.0</p><p>Copyright © 1982, 2026, Oracle and/or its affiliates. All rights reserved.</p><p>Welcome to DGMGRL, type “help” for information.<br>Password:<br>Connected to “cdb1”<br>Connected as SYSDBA.<br>DGMGRL&gt; show configuration</p><p>Configuration — my_dg_config</p><p>Protection Mode: MaxPerformance<br> Members:<br> cdb1 — Primary database<br> cdb1_stby — Physical standby database</p><p>Fast-Start Failover: Disabled</p><p>Configuration Status:<br>SUCCESS (status updated 54 seconds ago)</p><p>DGMGRL&gt;<br>DGMGRL&gt; VALIDATE DATABASE cdb1_stby;</p><p>Database Role: Physical standby database<br> Primary Database: cdb1</p><p>Ready for Switchover: Yes &lt;&lt;&lt;&lt; Ready to Switchover <br> Ready for Failover: Yes (Primary Running)</p><p>Flashback Database Status:<br> Database Status Retention Target<br> cdb1 On 1440<br> cdb1_stby Off 1440</p><p>Managed by Clusterware:<br> cdb1 : NO<br> cdb1_stby: NO<br> The static connect identifier allows for a connection to database “cdb1”.</p><p>Parameter Settings:<br> Parameter cdb1 Value cdb1_stby Value<br> DB_BLOCK_CHECKING FALSE FALSE<br> DB_BLOCK_CHECKSUM TYPICAL TYPICAL<br> DB_LOST_WRITE_PROTECT TYPICAL TYPICAL</p><p>DGMGRL&gt;<br>DGMGRL&gt; switchover to cdb1_stby;<br>2026–10–12T12:35:25.847+00:00<br>Performing switchover NOW, please wait…</p><p>2026–10–12T12:35:25.878+00:00<br>Operation requires a connection to database “cdb1_stby”<br>Connecting …<br>Connected to “cdb1_stby”<br>Connected as SYSDBA.</p><p>2026–10–12T12:35:26.297+00:00<br>Continuing with the switchover…</p><p>2026–10–12T12:35:31.277+00:00<br>New primary database “cdb1_stby” is opening…</p><p>2026–10–12T12:35:31.278+00:00<br>Operation requires start up of instance “cdb1” on database “cdb1”<br>Starting instance “cdb1”…<br>Connected to an idle instance.<br>ORACLE instance started.<br>Connected to “cdb1”<br>Database mounted.</p><p>2026–10–12T12:36:01.329+00:00<br>Switchover succeeded, new primary is “cdb1_stby”</p><p>2026–10–12T12:36:01.539+00:00<br>Switchover processing complete, broker ready.<br>DGMGRL&gt;</p><p>DGMGRL&gt; show configuration;</p><p>Configuration — my_dg_config</p><p>Protection Mode: MaxPerformance<br> Members:<br> cdb1_stby — Primary database<br> cdb1 — Physical standby database</p><p>Fast-Start Failover: Disabled</p><p>Configuration Status:<br>SUCCESS (status updated 40 seconds ago)</p><p>DGMGRL&gt;</p><p>DGMGRL&gt; show configuration</p><p>Configuration — my_dg_config</p><p>Protection Mode: MaxPerformance<br> Members:<br> cdb1 — Primary database<br> cdb1_stby — Physical standby database</p><p>Fast-Start Failover: Disabled</p><p>Configuration Status:<br>SUCCESS (status updated 16 seconds ago)</p><p>Converting physical standby to snapshot standby database</p><p>DGMGRL&gt; CONVERT DATABASE cdb1_stby to SNAPSHOT STANDBY;<br>2026–10–12T12:50:31.321+00:00<br>Converting database “cdb1_stby” to a Snapshot Standby database, please wait…</p><p>2026–10–12T12:50:43.552+00:00<br>Database “cdb1_stby” converted successfully</p><p>2026–10–12T12:50:43.552+00:00<br>DGMGRL&gt; show configuration</p><p>Configuration — my_dg_config</p><p>Protection Mode: MaxPerformance<br> Members:<br> cdb1 — Primary database<br> cdb1_stby — Snapshot standby database</p><p>Fast-Start Failover: Disabled</p><p>Configuration Status:<br>SUCCESS (status updated 42 seconds ago)</p><p>DGMGRL&gt;</p><p>SQL&gt; select EVENT,BEGIN_TIME,END_TIME FROM v$dg_broker_role_change;</p><p>EVENT BEGIN_TIME END_TIME<br> — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — <br>Switchover 12-OCT-26 10.55.55.000000000 AM 12-OCT-26 10.56.26.000000000 AM<br>Switchover 12-OCT-26 10.59.41.000000000 AM 12-OCT-26 11.00.02.000000000 AM<br>Switchover 12-OCT-26 11.01.46.000000000 AM 12-OCT-26 11.02.04.000000000 AM<br>Switchover 12-OCT-26 12.10.54.000000000 PM 12-OCT-26 12.11.23.000000000 PM<br>Switchover 12-OCT-26 12.35.26.000000000 PM 12-OCT-26 12.35.51.000000000 PM</p><p>SQL&gt;</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=30fbdd1f2f72" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Oracle AI Database 26ai True Cache]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-ai-database-26ai-true-cache-ae08e81b8848?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae08e81b8848</guid>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Sat, 03 Oct 2026 11:06:12 GMT</pubDate>
            <atom:updated>2026-10-03T11:06:12.788Z</atom:updated>
            <content:encoded><![CDATA[<h3>1. Overview</h3><p>Oracle True Cache is an application-transparent, read-only, transactionally-consistent mid-tier cache that sits between the application and the primary database. It is built on the same redo-apply technology as a physical standby: the primary ships redo to the True Cache instance, which continuously applies it, so cached data stays consistent without any application code to populate or invalidate the cache. The result removes the need to operate a separate Redis or Memcached tier.</p><p>Reads are routed to True Cache for low latency; writes continue to go to the primary. With the Oracle AI Database 26ai JDBC thin driver this is transparent — the application keeps the same connection URL and simply marks a connection read-only to route it to the cache.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*pzgzJi6ZylYxlPLBOTQk1Q.jpeg" /></figure><p><em>Figure 1 — Oracle True Cache read/write data flow</em></p><p><strong>About the figures: </strong>the console images in this guide are illustrative representations of the expected SQL*Plus and DBCA sessions, captured on a reference environment. Exact paths, SIDs, and timing on your system will differ — capture your own screens where local evidence is required.</p><h3>2. Prerequisites</h3><p>• Two hosts that can reach each other: the primary database node and a separate True Cache node (the cache may share the primary host for a lab, but a separate host is recommended for production).</p><p>• Oracle AI Database 26ai software installed on the True Cache node. The True Cache and primary editions must match (an Enterprise Edition True Cache requires an Enterprise Edition primary; Oracle Database Free supports at most one True Cache per Free primary).</p><p>• The primary database must run in ARCHIVELOG mode (Step 1).</p><p>• Do not pre-set LOG_ARCHIVE_CONFIG or LOG_ARCHIVE_DEST_n on the primary — True Cache configures these automatically.</p><p>• The primary password file (and the TDE wallet, if used) must be made available to the True Cache node — this guide packages them into a configuration BLOB with DBCA.</p><p>• A primary database application service for the data you want to cache must already exist before creating its True Cache service (Step 5).</p><h3>3. Step-by-Step Configuration</h3><h3>Step 1 — Enable ARCHIVELOG mode on the primary</h3><p>True Cache refreshes by applying redo, so the primary must be in ARCHIVELOG mode. Connect as SYSDBA, bounce the database into MOUNT, enable archiving, and reopen. Verify with ARCHIVE LOG LIST.</p><p>sqlplus / as sysdba<br> SHUTDOWN IMMEDIATE<br> STARTUP MOUNT<br> ALTER DATABASE ARCHIVELOG;<br> ALTER DATABASE OPEN;<br> ARCHIVE LOG LIST;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*fDABWvb52ZLxiWd25plO0Q.jpeg" /></figure><p><em>Figure 2 — console session: enabling ARCHIVELOG on the primary</em></p><p><strong>Important: </strong>do not set LOG_ARCHIVE_CONFIG or LOG_ARCHIVE_DEST_n yourself — True Cache will configure the archive destinations on the primary for you.</p><h3>Step 2 — Package credentials into a configuration BLOB</h3><p>On the primary, use DBCA to bundle the password file (and TDE wallet, if present) into a single configuration BLOB. This is simpler than copying each file separately.</p><p>$ORACLE_HOME/bin/dbca -configureDatabase \<br> -prepareTrueCacheConfigFile \<br> -sourceDB P26AI \<br> -trueCacheBlobLocation /home/oracle/scripts/ -silent</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*OVIA0nqGfTKfDC4TLeJDbQ.jpeg" /></figure><p><em>Figure 3 — console session: creating the True Cache configuration BLOB</em></p><h3>Step 3 — Copy the BLOB to the True Cache node</h3><p>Transfer the generated BLOB (and, if you chose not to use a BLOB, the password file and wallet) to a temporary location on the True Cache node.</p><p>scp /home/oracle/scripts/truecache_config.blob \<br> oracle@tc-host:/opt/oracle/tmp/</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*uUW0rCeg8hhM_ItJI2krVw.jpeg" /></figure><p><em>Figure 4 — console session: copying the BLOB to the cache node</em></p><h3>Step 4 — Create the True Cache instance</h3><p>On the True Cache node, run DBCA to create the True Cache instance, pointing it at the primary with an Easy Connect string and the configuration BLOB. DBCA verifies ARCHIVELOG mode, configures the archive destinations on the primary, and starts redo apply.</p><p>$ORACLE_HOME/bin/dbca -silent -createDatabase \<br> -databaseConfigType TRUECACHE \<br> -sourceDB “prim-host:1521/P26AI” \<br> -trueCacheConfigFile /opt/oracle/tmp/truecache_config.blob \<br> -gdbName FREE_TC -sid FREE_TC -silent</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*CB9zjtZk0fthmvkL967HyQ.jpeg" /></figure><p><em>Figure 5 — console session: creating the True Cache instance</em></p><p><strong>Verify the flags: </strong>the exact DBCA -createDatabase options for a True Cache instance vary by release and platform (and Oracle Database Free uses a simplified flow). Confirm against the Oracle True Cache User’s Guide 26ai before running in your environment.</p><h3>Step 5 — Create the True Cache application service</h3><p>Back on the primary, create the True Cache application service that corresponds to your existing primary read/write service. This is what lets applications switch reads to the cache without changing the connection URL. Run it once per service you want to cache.</p><p>$ORACLE_HOME/bin/dbca -configureDatabase \<br> -configureTrueCacheInstanceService \<br> -sourceDB P26AI \<br> -trueCacheConnectString “tc-host:1521/FREE_TC” \<br> -trueCacheServiceName sales_ro_tc \<br> -serviceName sales_rw \<br> -pdbName FREEPDB1 -silent</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*Dq6Onb8bzvjnynLH4JdMuQ.jpeg" /></figure><p><em>Figure 6 — console session: creating the True Cache application service</em></p><p><strong>Order matters: </strong>the primary application service (here sales_rw) must already exist before you run this command.</p><h3>Step 6 — Verify the True Cache instance</h3><p>On the True Cache node, confirm the instance is open READ ONLY WITH APPLY (redo is being applied) and that cached data is queryable.</p><p>sqlplus / as sysdba<br> SELECT open_mode FROM v$database;<br> SELECT COUNT(*) FROM sales.accounts;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*b23uOgm6LJ0Sck8ZW8VFqQ.jpeg" /></figure><p><em>Figure 7 — console session: verifying open mode and data on the cache</em></p><h3>Step 7 — Connect the application and test</h3><p>Point a read workload at the True Cache service and confirm it returns data. A write attempted through the True Cache service is correctly rejected, because the cache is read-only — writes must go to the primary service.</p><p>sqlplus sales_app@”tc-host:1521/sales_ro_tc”<br> SELECT balance FROM accounts WHERE id = 42;<br> UPDATE accounts SET balance = balance + 50 WHERE id = 42; — rejected</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/819/1*urKYJPnzIqygtQT_nZzYxQ.jpeg" /></figure><p><em>Figure 8 — console session: reading via True Cache; write is rejected (read-only)</em></p><h3>4. Application Integration (JDBC)</h3><p>The application-transparent design means the program keeps the same connection URL. With the 26ai JDBC thin driver, marking a connection read-only routes its reads to the corresponding True Cache service; clearing the flag routes work back to the primary.</p><p>// reads -&gt; True Cache (same URL, just mark the connection read-only)<br> conn.setReadOnly(true);<br> ResultSet rs = conn.createStatement()<br> .executeQuery(“SELECT balance FROM accounts WHERE id = 42”);<br> <br> // writes -&gt; primary<br> conn.setReadOnly(false);<br> conn.createStatement()<br> .executeUpdate(“UPDATE accounts SET balance = balance + 50 WHERE id = 42”);</p><p><strong>Staleness tolerance: </strong>True Cache reads are transactionally consistent but may lag the primary by the redo-apply interval. For read-your-own-writes paths, keep the connection on the primary; for read-mostly paths, the cache delivers the low-latency reads.</p><h3>5. Verification &amp; Troubleshooting</h3><p><strong>Check</strong></p><p><strong>Command / where</strong></p><p><strong>Expected</strong></p><p>Primary in ARCHIVELOG</p><p>ARCHIVE LOG LIST (primary)</p><p>Archive Mode / Enabled</p><p>Cache open mode</p><p>SELECT open_mode FROM v$database (cache)</p><p>READ ONLY WITH APPLY</p><p>Redo apply running</p><p>check archive dest on primary auto-configured</p><p>LOG_ARCHIVE_DEST_n points at the cache</p><p>Service registered</p><p>lsnrctl status (cache listener)</p><p>sales_ro_tc service present</p><p>Read works</p><p>SELECT via tc-host:1521/sales_ro_tc</p><p>rows returned</p><p>Write rejected</p><p>UPDATE via the cache service</p><p>ORA-16000 (read-only) — expected</p><p>Common issues: if DBCA reports the primary is not in ARCHIVELOG mode, repeat Step 1; if the cache cannot reach the primary, confirm the Easy Connect string and that the listener allows registration from the cache node (VNCR / REGISTRATION_INVITED_NODES); if authentication fails on the cache, re-check that the password file (or BLOB) was copied and matches the primary.</p><h3>6. Notes &amp; References</h3><p>• Oracle True Cache User’s Guide, Oracle AI Database 26ai — the authoritative source for prerequisites, exact DBCA parameters, and the manual (non-DBCA) configuration path.</p><p>• Oracle AI Database 26ai New Features Guide — True Cache overview and positioning.</p><p>• For a no-cost trial, Oracle AI Database 26ai Free supports a simplified True Cache configuration; a two-VM VirtualBox setup is a common way to practice this procedure end to end.</p><p><strong>GUI alternative: </strong>DBCA can also be driven through its graphical wizard. If you configure True Cache that way, capture the wizard’s progress and summary screens from your own environment and insert them in place of the corresponding console figures above.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae08e81b8848" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Oracle AI Database 26AI SQL Firewall]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-ai-database-26ai-sql-firewall-b83917c1a4fc?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/b83917c1a4fc</guid>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Wed, 30 Sep 2026 03:38:30 GMT</pubDate>
            <atom:updated>2026-09-30T03:38:30.330Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Oracle AI Database 26AI SQL Firewall</strong></p><p>Step-by-Step Implementation Guide</p><p>A practical runbook for enabling SQL Firewall with the DBMS_SQL_FIREWALL package: capturing a workload, generating an allow-list, enforcing it, and operating it over time.</p><h3>1. Overview</h3><p>SQL Firewall is built into the Oracle AI Database 26ai kernel. It inspects every incoming connection and SQL statement, including SQL issued from PL/SQL, whether local or over the network and whether encrypted or clear text. Only explicitly authorized SQL and connection paths are allowed; everything else can be logged or blocked. Because it runs inside the database, it cannot be bypassed.</p><p>Beyond SQL text, SQL Firewall can restrict how an account connects using session context: client IP address, OS user name and OS program name. It can be used in both the CDB root and pluggable databases (PDBs), and is designed for predictable application service accounts rather than SYS or ad-hoc DBA users.</p><p>The implementation follows four phases:</p><p><strong>Phase</strong></p><p><strong>Purpose</strong></p><p>1. Learn</p><p>Capture what the application account normally does.</p><p>2. Build</p><p>Generate an allow-list from the captured behavior.</p><p>3. Enforce</p><p>Apply the allow-list, first in observe mode, then blocking.</p><p>4. Operate</p><p>Monitor violations, tune the allow-list and maintain logs.</p><h3>2. Prerequisites</h3><h3>2.1 Licensing</h3><p>SQL Firewall is not part of the base Enterprise Edition license. It is included with the <strong>Oracle Database Vault</strong> option or with <strong>Oracle Audit Vault and Database Firewall</strong>. On Oracle Cloud Infrastructure, Database Vault is included at no additional cost on BaseDB EE-HP, BaseDB EE-EP and ExaDB-D. SQL Firewall is not available on Standard Edition.</p><p><strong>Note: </strong>Confirm licensing with the Oracle AI Database Licensing Information User Manual before enabling the feature in production.</p><h3>2.2 Roles</h3><p><strong>Role</strong></p><p><strong>Grants</strong></p><p>SQL_FIREWALL_ADMIN</p><p>Full administration through DBMS_SQL_FIREWALL.</p><p>SQL_FIREWALL_VIEWER</p><p>Read-only access to SQL Firewall views, for auditors and monitoring.</p><h3>3. Step-by-Step Implementation</h3><h3>Step 1: Grant the administration roles</h3><p>Connect to the target PDB as a privileged user and grant the roles:</p><p>GRANT SQL_FIREWALL_ADMIN TO sec_admin;</p><p>— Read-only access for auditors/monitoring:</p><p>GRANT SQL_FIREWALL_VIEWER TO auditor;</p><h3>Step 2: Enable SQL Firewall</h3><p>SQL Firewall is disabled by default. Enable it and confirm the status:</p><p>EXEC DBMS_SQL_FIREWALL.ENABLE;</p><p>SELECT status FROM DBA_SQL_FIREWALL_STATUS; — should show ENABLED</p><h3>Step 3: Start a capture for the application user</h3><p>A capture records the SQL statements and connection contexts the account normally uses.</p><p>BEGIN</p><p>DBMS_SQL_FIREWALL.CREATE_CAPTURE(</p><p>username =&gt; ‘APP_USER’,</p><p>top_level_only =&gt; FALSE, — also capture SQL issued inside PL/SQL</p><p>start_capture =&gt; TRUE);</p><p>END;</p><p>/</p><p>SELECT username, status FROM DBA_SQL_FIREWALL_CAPTURES;</p><h3>Step 4: Run a representative workload</h3><p>This is the most important step. Let the application run long enough to exercise all its normal paths:</p><p>• Daily transactions and user screens</p><p>• Batch jobs, month-end and quarter-end reports</p><p>• Administrative functions in the application</p><p>• Every application server, IP address and client program that connects</p><p><strong>Tip: </strong>Anything not captured here will be flagged later. Many teams capture for one to four weeks in a test or pre-production environment that mirrors production traffic.</p><h3>Step 5: Stop the capture and review what was learned</h3><p>EXEC DBMS_SQL_FIREWALL.STOP_CAPTURE(‘APP_USER’);</p><p>— SQL captured</p><p>SELECT command_type, sql_text, current_user, top_level</p><p>FROM DBA_SQL_FIREWALL_CAPTURE_LOGS</p><p>WHERE username = ‘APP_USER’;</p><p>— Connection contexts captured</p><p>SELECT ip_address, client_program, os_user</p><p>FROM DBA_SQL_FIREWALL_SESSION_LOGS</p><p>WHERE username = ‘APP_USER’;</p><p>Check for anything that should not be allowed, such as a developer laptop IP or a one-off manual query, before building the allow-list.</p><h3>Step 6: Generate the allow-list</h3><p>EXEC DBMS_SQL_FIREWALL.GENERATE_ALLOW_LIST(‘APP_USER’);</p><p>SELECT * FROM DBA_SQL_FIREWALL_ALLOW_LISTS WHERE username = ‘APP_USER’;</p><p>SELECT * FROM DBA_SQL_FIREWALL_ALLOWED_SQL WHERE username = ‘APP_USER’;</p><p>SELECT * FROM DBA_SQL_FIREWALL_ALLOWED_IP_ADDR WHERE username = ‘APP_USER’;</p><p>SELECT * FROM DBA_SQL_FIREWALL_ALLOWED_OS_PROG WHERE username = ‘APP_USER’;</p><p>SELECT * FROM DBA_SQL_FIREWALL_ALLOWED_OS_USER WHERE username = ‘APP_USER’;</p><h3>Step 7: Enable the allow-list in observe mode</h3><p>Start with block =&gt; FALSE. Violations are logged but not stopped, so gaps can be found without breaking the application.</p><p>BEGIN</p><p>DBMS_SQL_FIREWALL.ENABLE_ALLOW_LIST(</p><p>username =&gt; ‘APP_USER’,</p><p>enforce =&gt; DBMS_SQL_FIREWALL.ENFORCE_ALL, — or ENFORCE_SQL</p><p>block =&gt; FALSE);</p><p>END;</p><p>/</p><p>Enforcement modes:</p><p><strong>Mode</strong></p><p><strong>What is checked</strong></p><p>ENFORCE_CONTEXT</p><p>Only IP address, OS program and OS user</p><p>ENFORCE_SQL</p><p>Only whether the SQL statement is in the allow-list</p><p>ENFORCE_ALL</p><p>Both connection context and SQL</p><h3>Step 8: Review violations and tune</h3><p>SELECT occurred_at, cause, firewall_action, sql_text,</p><p>ip_address, client_program, os_user</p><p>FROM DBA_SQL_FIREWALL_VIOLATIONS</p><p>WHERE username = ‘APP_USER’</p><p>ORDER BY occurred_at DESC;</p><p>The CAUSE column shows whether the violation was a context violation or a SQL violation; FIREWALL_ACTION shows whether it was allowed or blocked.</p><p>To add legitimate items missed during capture:</p><p>— Add a new application server IP</p><p>BEGIN</p><p>DBMS_SQL_FIREWALL.ADD_ALLOWED_CONTEXT(</p><p>username =&gt; ‘APP_USER’,</p><p>context_type =&gt; DBMS_SQL_FIREWALL.IP_ADDRESS,</p><p>value =&gt; ‘10.20.30.40’);</p><p>END;</p><p>/</p><p>— Append all logged violations to the allow-list (only after reviewing them)</p><p>BEGIN</p><p>DBMS_SQL_FIREWALL.APPEND_ALLOW_LIST(</p><p>username =&gt; ‘APP_USER’,</p><p>source =&gt; DBMS_SQL_FIREWALL.VIOLATION_LOG);</p><p>END;</p><p>/</p><p><strong>New in 26ai: </strong>APPEND_ALLOW_LIST_SINGLE_SQL adds one specific statement to the allow-list. Previously the whole violation or capture log had to be appended and unwanted entries deleted manually. Check the DBMS_SQL_FIREWALL reference for its exact parameters.</p><p>Remove entries that should not be allowed with DELETE_ALLOWED_CONTEXT or DELETE_ALLOWED_SQL.</p><h3>Step 9: Switch to blocking mode</h3><p>Once remaining violations are only genuine anomalies, turn blocking on:</p><p>BEGIN</p><p>DBMS_SQL_FIREWALL.UPDATE_ALLOW_LIST_ENFORCEMENT(</p><p>username =&gt; ‘APP_USER’,</p><p>enforce =&gt; DBMS_SQL_FIREWALL.ENFORCE_ALL,</p><p>block =&gt; TRUE);</p><p>END;</p><p>/</p><p>Blocked statements now fail with an ORA-47605 SQL Firewall violation error, and each attempt is still recorded in the violations view.</p><h3>Step 10: Test the protection</h3><p>From an unapproved client, or with a query the application never runs:</p><p>— As APP_USER:</p><p>SELECT * FROM app_owner.customers WHERE 1=1 OR ‘a’=’a’;</p><p>— Expected: ORA-47605 and a new row in DBA_SQL_FIREWALL_VIOLATIONS</p><h3>4. Ongoing Operations</h3><h3>4.1 Purge old logs</h3><p>BEGIN</p><p>DBMS_SQL_FIREWALL.PURGE_LOG(</p><p>username =&gt; ‘APP_USER’,</p><p>purge_time =&gt; SYSTIMESTAMP — INTERVAL ‘30’ DAY,</p><p>log_type =&gt; DBMS_SQL_FIREWALL.VIOLATION_LOG);</p><p>END;</p><p>/</p><h3>4.2 Application releases</h3><p>After each release, start a new capture while the allow-list stays in observe mode, run the new functionality, then append the new SQL to the allow-list before re-enabling blocking.</p><h3>4.3 Auditing and alerting</h3><p>Integrate with Unified Auditing so violations reach your SIEM, and alert on any blocked action.</p><h3>4.4 Fleet management</h3><p>Oracle Data Safe can administer and monitor SQL Firewall across many 26ai databases from one console, which is easier than running PL/SQL on each PDB.</p><h3>4.5 Disabling or removing SQL Firewall</h3><p>EXEC DBMS_SQL_FIREWALL.DISABLE_ALLOW_LIST(‘APP_USER’); — one user</p><p>EXEC DBMS_SQL_FIREWALL.DROP_ALLOW_LIST(‘APP_USER’);</p><p>EXEC DBMS_SQL_FIREWALL.DROP_CAPTURE(‘APP_USER’);</p><p>EXEC DBMS_SQL_FIREWALL.DISABLE; — all users</p><h3>5. Common Pitfalls</h3><p><strong>Pitfall</strong></p><p><strong>Recommendation</strong></p><p>Capture period too short</p><p>Stay in observe mode through at least one full business cycle so month-end and quarterly jobs are captured.</p><p>Dynamic SQL with literals</p><p>Applications that build SQL through string concatenation create many distinct statements. Bind variables make allow-lists far more stable.</p><p>Changing IPs (connection pools, scaling, Kubernetes)</p><p>Use ENFORCE_SQL alone if client IPs change frequently.</p><p>Applying it to DBA accounts</p><p>Use it for predictable application accounts; human users run too much varied SQL.</p><h3>6. Key Views Reference</h3><p><strong>View</strong></p><p><strong>Contents</strong></p><p>DBA_SQL_FIREWALL_STATUS</p><p>Whether SQL Firewall is enabled</p><p>DBA_SQL_FIREWALL_CAPTURES</p><p>Captures and their status</p><p>DBA_SQL_FIREWALL_CAPTURE_LOGS</p><p>SQL recorded during capture</p><p>DBA_SQL_FIREWALL_SESSION_LOGS</p><p>Connection contexts recorded</p><p>DBA_SQL_FIREWALL_ALLOW_LISTS</p><p>Allow-lists and enforcement settings</p><p>DBA_SQL_FIREWALL_ALLOWED_SQL</p><p>Allowed SQL statements</p><p>DBA_SQL_FIREWALL_ALLOWED_IP_ADDR / _OS_PROG / _OS_USER</p><p>Allowed connection contexts</p><p>DBA_SQL_FIREWALL_VIOLATIONS</p><p>Logged and blocked violations</p><h3>7. References</h3><p>• <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/nfcoa/sql_firewall.html">SQL Firewall — Oracle AI Database 26ai New Features</a></p><p>• <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/sqlfw/changes-this-release-oracle-database-sql-firewall-guide.html">Changes in This Release for Oracle AI Database SQL Firewall Guide</a></p><p>• <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/refrn/DBA_SQL_FIREWALL_VIOLATIONS.html">DBA_SQL_FIREWALL_VIOLATIONS reference</a></p><p>• <a href="https://docs.oracle.com/en-us/iaas/data-safe/doc/sql-firewall-overview.html">SQL Firewall Overview — Oracle Data Safe</a></p><p>• <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/sqlfw/licensing-oracle-sql-firewall.html">Licensing Oracle SQL Firewall</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b83917c1a4fc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Oracle 26 AI Database Local Rolling Maintenance & Smart Connection Rebalancing]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-26-ai-database-local-rolling-maintenance-smart-connection-rebalancing-e3856d4a5684?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/e3856d4a5684</guid>
            <category><![CDATA[oracle-26ai]]></category>
            <category><![CDATA[real-application-cluster]]></category>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 08:21:52 GMT</pubDate>
            <atom:updated>2026-09-25T08:21:52.974Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>ORACLE AI DATABASE 26ai</strong></p><p><strong>Local Rolling Maintenance &amp; Smart Connection Rebalancing</strong></p><p><em>A Step-by-Step Implementation Guide for DBAs</em></p><p>Continuous availability features for Oracle Real Application Clusters (RAC) and RAC One Node</p><p>Prepared for: Database Administration Team</p><p>Release: Oracle AI Database 26ai | Document version 1.0</p><h3>Contents</h3><ol><li><a href="#_Toc241235976">Contents</a></li><li><a href="#_Toc241235977">Introduction</a></li><li><a href="#_Toc241235978">The Two Features at a Glance</a></li><li><a href="#_Toc241235979">Local Rolling Database Maintenance</a></li><li><a href="#_Toc241235980">What it does</a></li><li><a href="#_Toc241235981">Prerequisites</a></li><li><a href="#_Toc241235982">Step-by-step procedure</a></li><li><a href="#_Toc241235983">Verifying the maintenance</a></li><li><a href="#_Toc241235984">Rolling back</a></li><li><a href="#_Toc241235985">Smart Connection Rebalancing</a></li><li><a href="#_Toc241235986">What it does</a></li><li><a href="#_Toc241235987">How it fits the runtime stack</a></li><li><a href="#_Toc241235988">Step-by-step configuration</a></li><li><a href="#_Toc241235989">Verifying rebalancing behaviour</a></li><li><a href="#_Toc241235990">Single-Instance Databases: Single-Server Rolling Maintenance</a></li><li><a href="#_Toc241235991">Best Practices &amp; Troubleshooting</a></li><li><a href="#_Toc241235992">Best practices</a></li><li><a href="#_Toc241235993">Common issues</a></li><li><a href="#_Toc241235994">References</a></li></ol><h3>1. Introduction</h3><p>Oracle AI Database 26ai shifts the emphasis of the release toward continuous operation: the expectation that a production platform stays online while it is being patched, upgraded, rebalanced, and recovered. Two features at the centre of that goal are Local Rolling Database Maintenance and Smart Connection Rebalancing. This guide explains what each feature does, the prerequisites involved, and a practical, step-by-step procedure to implement and verify them.</p><p><strong>Local Rolling Database Maintenance</strong> lets you apply patches and perform maintenance on an Oracle RAC or RAC One Node database in out-of-place mode, locally on a node, without interrupting the database workload. <strong>Smart Connection Rebalancing</strong> transparently redistributes service-based sessions across surviving and returning instances based on real-time performance, so the cluster settles quickly after any topology change — including the brief two-instance window created by rolling maintenance.</p><p><strong>Audience </strong>DBAs, platform engineers, and architects responsible for Oracle RAC / RAC One Node estates running Oracle AI Database 26ai. Familiarity with SRVCTL, Grid Infrastructure, and out-of-place (gold image) patching is assumed.</p><p><strong>Important </strong>Command syntax and flags evolve across Release Updates. Always confirm exact options against the patch README and the official 26ai documentation for your platform before running anything in production.</p><h3>2. The Two Features at a Glance</h3><p><strong>Aspect</strong></p><p><strong>Local Rolling Maintenance</strong></p><p><strong>Smart Connection Rebalancing</strong></p><p>Purpose</p><p>Patch / maintain an instance out-of-place on its own node without downtime</p><p>Automatically move sessions to the best-performing instances after change</p><p>Scope</p><p>RAC and RAC One Node (single instance via Single-Server Rolling Maintenance)</p><p>Service-based connections across RAC instances</p><p>Trigger</p><p>Planned: DBA-initiated patching or maintenance</p><p>Automatic: real-time performance monitoring and topology events</p><p>Key benefit</p><p>Keeps workload local; reduces brownout time during patching</p><p>Smooth workload redistribution; no manual session juggling</p><p>Admin effort</p><p>Procedural — you orchestrate the steps (or use Oracle FPP)</p><p>Largely hands-off once services and Application Continuity are configured</p><p><strong>How they complement each other </strong>Rolling maintenance temporarily runs two instances of the same database on a node and drains the old one. Smart Connection Rebalancing is what makes that drain graceful — it reshuffles sessions onto healthy instances instead of leaving the DBA to migrate connections by hand.</p><h3>3. Local Rolling Database Maintenance</h3><h3>3.1 What it does</h3><p>Local Rolling Database Maintenance creates a new, separately patched (out-of-place) database home on the same server and starts a second instance of the same database from that new home. Sessions are drained from the instance running on the old home to the instance on the new home, the old instance is shut down, and the maintenance is complete — all without taking the database offline. Because the workload stays on the local node, the brownout window is shorter than approaches that relocate the workload to another node, and the impact on the rest of the cluster is limited.</p><p>• Applies to Oracle RAC and Oracle RAC One Node databases.</p><p>• Runs in out-of-place mode — the patch goes into a new home, leaving the original untouched and available for rollback.</p><p>• Can be orchestrated manually with custom scripts or automated with Oracle Fleet Patching and Provisioning (FPP), which in 26ai includes a local-mode option that runs on the database server itself.</p><h3>3.2 Prerequisites</h3><p>1. <strong>RAC or RAC One Node deployment. </strong>The database must be Oracle RAC or RAC One Node on a Grid Infrastructure cluster. A pure single-instance database must first be converted (see Section 5).</p><p>2. <strong>Spare capacity for two instances. </strong>The node must have enough CPU, memory, and other resources to run two instances of the database simultaneously during the transition.</p><p>3. <strong>A patched (gold image) Oracle home. </strong>Prepare the new, out-of-place database home with the target Release Update or one-off patch applied.</p><p>4. <strong>Drainable services. </strong>Application services should be configured to drain (and, ideally, use Application Continuity / Transparent Application Continuity) so sessions migrate cleanly.</p><p>5. <strong>Readiness check passed. </strong>Validate the server is ready for local rolling patching before you begin.</p><p><em>Run the cluster verification readiness check:</em></p><p>cluvfy stage -pre patch -db -dbhome &lt;oracle_home&gt; \ -dest_dbhome &lt;dest_dbhome&gt; -localrolling</p><h3>3.3 Step-by-step procedure</h3><p><strong>Tip </strong>Take a fresh RMAN backup and capture current service / instance configuration (srvctl config database -d &lt;db&gt;) before you start, so you have a known-good baseline.</p><p>1. <strong>Stage the new Oracle home. </strong>Install the new database software out-of-place and apply the target patch (RU or one-off), creating &lt;dest_dbhome&gt;.</p><p>2. <strong>Run the readiness check. </strong>Execute the cluvfy command shown in 3.2 and resolve any reported gaps before continuing.</p><p>3. <strong>Initiate local rolling maintenance. </strong>Start a second instance of the database from the new home on the same node. This is the patched instance that will take over the workload.</p><p>4. <strong>Drain and migrate sessions. </strong>Relocate the service so sessions drain from the old-home instance to the new-home instance. Smart Connection Rebalancing and service draining handle the redistribution gracefully.</p><p>5. <strong>Shut down the old-home instance. </strong>Once sessions have drained, stop the instance that is still running from the original home.</p><p>6. <strong>Confirm the database now runs from the new home. </strong>The database is now serving the full workload from the patched home; the original home remains intact for rollback.</p><p>7. <strong>Repeat per node. </strong>For a multi-node RAC database, repeat the procedure on each remaining node, one at a time, so the database is never fully offline.</p><p><strong>Automation </strong>Oracle FPP can orchestrate all of the above with the rhpctl move workflow and, in 26ai, a local-mode option that does not require a separate FPP server. This is the recommended method for fleets and for RAC + Data Guard estates.</p><h3>3.4 Verifying the maintenance</h3><p>After the workload is running from the patched home, confirm the configuration and patch level:</p><p>— Confirm where the database is running from<br> srvctl config database -d &lt;db_unique_name&gt;<br> srvctl status database -d &lt;db_unique_name&gt;<br> <br> — Confirm patch / RU level<br> $ORACLE_HOME/OPatch/opatch lspatches</p><p>• All target instances report the new Oracle home.</p><p>• opatch lspatches lists the expected Release Update or one-off patch.</p><p>• Application services are running and accepting connections on the patched instances.</p><h3>3.5 Rolling back</h3><p>Because maintenance is out-of-place, rollback means moving the database back to the original home. With FPP this is the reverse rhpctl move (swapping the source and destination homes). Performing the move in rolling fashion keeps the database available during the rollback as well.</p><p>rhpctl move database -sourcehome &lt;new_home&gt; \ -desthome &lt;old_home&gt; -dbname &lt;db&gt; -localrolling</p><h3>4. Smart Connection Rebalancing</h3><h3>4.1 What it does</h3><p>Smart Connection Rebalancing transparently reshuffles service-based connections across Oracle RAC instances based on real-time performance monitoring. Instead of relying purely on connect-time load balancing, the database continuously evaluates instance performance and moves sessions to where they will run best — without manual intervention. It works hand-in-hand with smooth reconfiguration, which reduces the brownout caused by RAC operations such as nodes joining or leaving the cluster.</p><p>• Operates on service-based connections across RAC instances.</p><p>• Driven by real-time performance metrics, not just static or round-robin balancing.</p><p>• Reacts automatically to topology changes — the very events rolling maintenance produces.</p><h3>4.2 How it fits the runtime stack</h3><p>Rebalancing is most effective when the surrounding connection and high-availability features are in place. The pieces work together as a layered stack:</p><p><strong>Layer</strong></p><p><strong>Role in smooth rebalancing</strong></p><p>Database services</p><p>The unit of work that is balanced and relocated. Define services per workload rather than connecting to the database name directly.</p><p>Runtime load balancing</p><p>Distributes new and existing work toward better-performing instances using real-time advisory data.</p><p>Draining</p><p>Lets in-flight work finish on an instance being vacated, so sessions move without errors.</p><p>Application Continuity / TAC</p><p>Replays in-flight transactions on the new instance so users do not see interruptions, even for sticky sessions.</p><p>Smooth reconfiguration</p><p>Reduces the brownout window when instances join or leave, so rebalancing settles quickly.</p><h3>4.3 Step-by-step configuration</h3><p><strong>Goal </strong>Get connections onto well-defined services with runtime load balancing, draining, and Application Continuity enabled, so Smart Connection Rebalancing has everything it needs to move sessions transparently.</p><p>1. <strong>Create (or review) application services. </strong>Connect applications through services, not the raw database name. Set a goal for runtime load balancing and a non-zero drain timeout.</p><p>srvctl add service -d &lt;db_unique_name&gt; -s &lt;service_name&gt; \<br> -preferred &lt;inst1,inst2&gt; -available &lt;inst3&gt; \<br> -rlbgoal SERVICE_TIME -clbgoal SHORT \<br> -failovertype AUTO -failovermethod BASIC \<br> -drain_timeout 120 -stopoption IMMEDIATE \<br> -commit_outcome TRUE</p><p>2. <strong>Enable runtime load balancing in the client pool. </strong>Use a connection pool (UCP, ODP.NET, node-oracledb, etc.) that consumes Fast Application Notification (FAN) and runtime load-balancing advisories so it can act on rebalancing events.</p><p>3. <strong>Use Application Continuity or Transparent Application Continuity. </strong>This allows in-flight transactions to be replayed on the target instance, making session moves invisible to users.</p><p>4. <strong>Confirm FAN is being delivered. </strong>Clients must receive FAN events for draining and rebalancing to be graceful. Verify the connection string uses the SCAN and that ONS / AQ-HA notifications reach the pool.</p><p>5. <strong>Start the service and let rebalancing operate. </strong>Once services are running with these settings, Smart Connection Rebalancing redistributes sessions automatically as performance and topology change — no per-event action required.</p><p>srvctl start service -d &lt;db_unique_name&gt; -s &lt;service_name&gt; srvctl status service -d &lt;db_unique_name&gt; -s &lt;service_name&gt;</p><h3>4.4 Verifying rebalancing behaviour</h3><p>To confirm sessions are distributed and moving as expected, inspect session counts per instance and watch them rebalance during a topology change (for example, the two-instance window of a rolling patch):</p><p>— Sessions per instance for a given service<br> SELECT inst_id, COUNT(*) AS sessions<br> FROM gv$session<br> WHERE service_name = ‘&lt;service_name&gt;’<br> GROUP BY inst_id<br> ORDER BY inst_id;</p><p>• Session counts are spread across the preferred instances rather than piling onto one.</p><p>• During an instance add/remove, counts shift toward the healthy instances without application errors.</p><p>• Connection-pool metrics show borrowed connections following the advisory, not erroring out.</p><h3>5. Single-Instance Databases: Single-Server Rolling Maintenance</h3><p>If your database is plain single-instance (not RAC or RAC One Node), you can still benefit from rolling maintenance through the Single-Server Rolling Database Maintenance feature. The approach converts the single instance into a RAC One Node database on a single-node cluster, then applies the same local rolling patching workflow.</p><p><strong>Precondition </strong>Conversion is possible when the single-instance database uses file systems accessible to the single server. Run the cluvfy readiness check (Section 3.2) before converting.</p><p>1. <strong>Convert single instance to RAC One Node. </strong>Use Oracle DBCA to convert the single-instance database to a RAC One Node database on the single-node cluster.</p><p>2. <strong>Create the new patched home. </strong>Stage the out-of-place database home with the target patch applied.</p><p>3. <strong>Enable single-server rolling maintenance. </strong>This creates and starts a new instance of the same database from the new home on the same server.</p><p>4. <strong>Transfer the database to the new home. </strong>Drain to the new-home instance and shut down the old one, following the local rolling patching procedure in Section 3.3.</p><p>The steps can be orchestrated with custom scripts or with Oracle FPP, the same as for multi-node RAC.</p><h3>6. Best Practices &amp; Troubleshooting</h3><h3>6.1 Best practices</h3><p>• Always patch out-of-place into a gold image; keep the previous home for fast rollback.</p><p>• Size nodes for the brief two-instance window — verify CPU and memory headroom before starting.</p><p>• Drive all client connections through services with draining and Application Continuity enabled.</p><p>• Patch one node at a time and validate before moving to the next.</p><p>• Use FPP for fleets and for RAC + Data Guard estates to reduce manual error.</p><p>• Test the full procedure in a non-production cluster that mirrors production topology first.</p><h3>6.2 Common issues</h3><p><strong>Symptom</strong></p><p><strong>Likely cause / action</strong></p><p>Sessions error during drain</p><p>FAN not reaching the pool, or no Application Continuity. Verify SCAN/ONS notifications and enable AC/TAC and a sensible drain_timeout.</p><p>Node runs out of resources</p><p>Insufficient headroom for two instances. Re-check CPU/memory; consider patching at a lower-load window.</p><p>cluvfy readiness check fails</p><p>Resolve every reported item before proceeding; do not bypass the pre-patch check.</p><p>Sessions pile on one instance</p><p>Runtime load balancing not active in the client pool, or service goals not set. Set -rlbgoal/-clbgoal and use a FAN-aware pool.</p><h3>7. References</h3><p>The following Oracle AI Database 26ai documentation pages cover these features in detail. Verify exact command syntax against the version that matches your environment.</p><p>• Single-Server Rolling Database Maintenance — Database Installation Guide (26ai) — <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/ladbi/single-server-rolling-database-maintenance.html">https://docs.oracle.com/en/database/oracle/oracle-database/26/ladbi/single-server-rolling-database-maintenance.html</a></p><p>• Administering Database Instances and Cluster Databases — RAC Administration and Deployment Guide (26ai) — <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/racad/administering-database-instances-and-cluster-databases.html">https://docs.oracle.com/en/database/oracle/oracle-database/26/racad/administering-database-instances-and-cluster-databases.html</a></p><p>• Oracle AI Database New Features — Continuous Availability (26ai) — <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/nfcoa/co_continuous_availability.html">https://docs.oracle.com/en/database/oracle/oracle-database/26/nfcoa/co_continuous_availability.html</a></p><p>• Oracle AI Database New Features — High Availability / RAC (26ai) — <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/nfcoa/ha_rac.html">https://docs.oracle.com/en/database/oracle/oracle-database/26/nfcoa/ha_rac.html</a></p><p>• Recommended Patch Maintenance for Oracle RAC (26ai) — <a href="https://docs.oracle.com/en/database/oracle/oracle-database/26/dbprc/">https://docs.oracle.com/en/database/oracle/oracle-database/26/dbprc/</a></p><p><strong>Disclaimer: </strong><em>This guide is a practitioner-oriented summary compiled from Oracle 26ai documentation and is not a substitute for the official Oracle product documentation. Validate all procedures in a test environment before applying them to production.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e3856d4a5684" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Oracle Cerifications]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-cerifications-57c383b28442?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/57c383b28442</guid>
            <category><![CDATA[oracle-certification]]></category>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 16:00:47 GMT</pubDate>
            <atom:updated>2026-09-17T16:00:47.214Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Wq2QDpHuTP13UNAlxQRS3Q.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yb2S7cJLHM14Musdh2T7ag.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/552/1*C8kDKJBbJJ30LXu5Mj7kag.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/552/1*TaIgPpDyQyu9-WWX2SN7Cg.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=57c383b28442" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Oracle GoldenGate Micro services Hub Setup with High Availability]]></title>
            <link>https://medium.com/@ashoka.dba/oracle-goldengate-micro-services-hub-setup-with-high-availability-dd174833dc53?source=rss-05f8b1c04f1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/dd174833dc53</guid>
            <category><![CDATA[micro-services-hub]]></category>
            <category><![CDATA[high-availability]]></category>
            <category><![CDATA[oracle-goldengate]]></category>
            <dc:creator><![CDATA[Ashoka Reddy Tatiparthi]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 14:01:37 GMT</pubDate>
            <atom:updated>2026-09-17T14:29:21.950Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Step-by-Step Implementation Guide</em></p><p>Architecture: Micro services Architecture (MA) Hub</p><p>• Oracle Grid Infrastructure + XAG + ACFS + NGINX</p><p>Target release: Oracle GoldenGate 23ai MA (steps also apply to 21c</p><h3>1. About This Document</h3><p><strong>Purpose. </strong>This document is a practical, step-by-step run book for deploying an Oracle GoldenGate Micro services Architecture (MA) “Hub” that is highly available. In a Hub topology the GoldenGate software runs off-box on a dedicated cluster rather than on the database servers, and a single centralized deployment connects remotely to one or more source and target databases.</p><p><strong>High-availability model. </strong>Availability is provided by Oracle Grid Infrastructure (Cluster ware) together with the Oracle Grid Infrastructure Standalone Agent (XAG). XAG manages the dependencies between the shared file system (ACFS), an application VIP, the NGINX reverse proxy, and the GoldenGate deployment, and relocates the whole stack between cluster nodes on failure or planned maintenance.</p><p><strong>Assumptions — </strong>Two-node Oracle Grid Infrastructure cluster on Oracle Linux 8/9; Oracle GoldenGate 23ai Micro services; ASM available for ACFS; XAG 10.2 or later; source and target databases already configured for MAA (RAC and/or Data Guard with role-based services). Adjust host names, IPs, paths, ports, and sizes to match your environment. Commands are examples — always confirm exact syntax with your installed version using the tool’s — help.</p><p><strong>Audience. </strong>Database administrators and platform engineers responsible for replication infrastructure.</p><h3>Table of Contents</h3><ol><li><a href="#_Toc239846455"><em>About This Document</em></a></li><li><a href="#_Toc239846456">Table of Contents</a></li><li><a href="#_Toc239846457">Architecture Overview</a></li><li><a href="#_Toc239846458">Core components</a></li><li><a href="#_Toc239846459">How a failover works</a></li><li><a href="#_Toc239846460">Reference environment used in this guide</a></li><li><a href="#_Toc239846461">Default microservice ports (bound to localhost; fronted by NGINX)</a></li><li><a href="#_Toc239846462">Prerequisites Checklist</a></li><li><a href="#_Toc239846463">Task 1 — Configure ACFS Shared Storage</a></li><li><a href="#_Toc239846464">Create the ASM volume and ACFS file system (as the grid user)</a></li><li><a href="#_Toc239846465">Register the file system with Clusterware (as root/grid)</a></li><li><a href="#_Toc239846466">Task 2 — Create the Application VIP</a></li><li><a href="#_Toc239846467">Task 3 — Install GoldenGate MA Software on Every Node</a></li><li><a href="#_Toc239846468">Interactive install</a></li><li><a href="#_Toc239846469">Silent install (repeat identically on each node)</a></li><li><a href="#_Toc239846470">Task 4 — Create the Microservices Deployment on ACFS</a></li><li><a href="#_Toc239846471">Silent creation (optional)</a></li><li><a href="#_Toc239846472">Task 5 — Install and Configure XAG</a></li><li><a href="#_Toc239846473">Task 6 — Register the Deployment with XAG</a></li><li><a href="#_Toc239846474">Start and verify</a></li><li><a href="#_Toc239846475">Task 7 — Configure the NGINX Reverse Proxy</a></li><li><a href="#_Toc239846476">Install NGINX</a></li><li><a href="#_Toc239846477">Generate the proxy configuration</a></li><li><a href="#_Toc239846478">Install a certificate, validate, and reload</a></li><li><a href="#_Toc239846479">Make NGINX cluster-managed and add it to the XAG dependency</a></li><li><a href="#_Toc239846480">Task 8 — Prepare Source and Target Databases (MAA)</a></li><li><a href="#_Toc239846481">Enable logging and create the GoldenGate user</a></li><li><a href="#_Toc239846482">Create role-based services (they run only on the primary)</a></li><li><a href="#_Toc239846483">Add TNS entries on the hub pointing at those services</a></li><li><a href="#_Toc239846484">Task 9 — Configure Replication (Extract, Path, Replicat)</a></li><li><a href="#_Toc239846485">Connect and set up credentials</a></li><li><a href="#_Toc239846486">Create the Integrated Extract (source)</a></li><li><a href="#_Toc239846487">Create a Distribution Path (hub to hub / to receiver)</a></li><li><a href="#_Toc239846488">Create the Integrated Replicat (target)</a></li><li><a href="#_Toc239846489">Task 10 — Validate High Availability</a></li><li><a href="#_Toc239846490">Planned relocation (maintenance test)</a></li><li><a href="#_Toc239846491">Checks after relocation</a></li><li><a href="#_Toc239846492">Unplanned failover test</a></li><li><a href="#_Toc239846493">Database switchover test</a></li><li><a href="#_Toc239846494">Monitoring, Maintenance, and Patching</a></li><li><a href="#_Toc239846495">Troubleshooting Quick Reference</a></li><li><a href="#_Toc239846496">Appendix A — Common agctl Commands</a></li><li><a href="#_Toc239846497">Appendix B — Build Order Summary</a></li></ol><h3>2. Architecture Overview</h3><p>A GoldenGate MA Hub separates the replication engine from the databases it serves. The benefits are centralized management, simpler patching, and the ability to protect the replication tier independently of the databases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*D6-Hpe4tL_LhFBdH4zfMnw.png" /></figure><h3>2.1 Core components</h3><p>• <strong>Oracle Grid Infrastructure / Clusterware — </strong>provides the cluster framework, VIP management, and ACFS.</p><p>• <strong>ACFS (ASM Cluster File System) — </strong>shared storage that holds the GoldenGate deployment home and trail files, so the same data is visible after relocation.</p><p>• <strong>Application VIP — </strong>a floating IP that always points to the node currently running the deployment; clients and the reverse proxy use it.</p><p>• <strong>Oracle GoldenGate MA software — </strong>installed identically on every node in a local (non-shared) home. Services: Service Manager, Administration Service, Distribution Service, Receiver Service, and Performance Metrics Service.</p><p>• <strong>XAG (Grid Infrastructure Standalone Agent) — </strong>registers the deployment as a Clusterware resource and coordinates start/stop/relocate across the ACFS mount, VIP, and NGINX.</p><p>• <strong>NGINX reverse proxy — </strong>presents the microservices behind a single hardened HTTPS port (443) and terminates TLS.</p><p>• <strong>MAA databases — </strong>source and target databases use RAC and/or Data Guard with role-based database services, so Extract and Replicat automatically follow the primary via the Oracle Net connect string.</p><h3>2.2 How a failover works</h3><p>1. Clusterware detects the node or resource failure.</p><p>2. XAG stops the deployment stack (if still up) on the failed node.</p><p>3. The ACFS mount, application VIP, and NGINX are brought online on the surviving node.</p><p>4. XAG starts the GoldenGate deployment; Extract and Replicat resume from the last checkpoint on the shared trail.</p><p>5. Because the VIP moved with the deployment, clients and the reverse proxy reconnect without configuration changes.</p><h3>2.3 Reference environment used in this guide</h3><p><strong>Item Example value</strong></p><p>Cluster nodes ggnode1, ggnode2 (2-node GI cluster)</p><p>Application VIP (host / IP) gghub-vip / 10.10.10.50</p><p>ACFS mount point /mnt/acfs_gg</p><p>Golden Gate software home (local, each node) /u01/app/oracle/goldengate/gg23ai</p><p>Deployment home (on ACFS) /mnt/acfs_gg/deployments/gghub1</p><p>Deployment name gghub1</p><p>XAG home (outside GI home) /u01/app/grid/xag</p><p>GoldenGate admin user oggadmin</p><p>External access URL <a href="https://gghub-vip/">https://gghub-vip:443</a></p><h3>Default microservice ports (bound to localhost; fronted by NGINX)</h3><p><strong>Service Example port Exposure</strong></p><p>NGINX reverse proxy (HTTPS) 443</p><p>External (via VIP) Service Manager 9100 localhost only</p><p>Administration Service 9101 localhost only</p><p>Distribution Service 9102 localhost only</p><p>Receiver Service 9103 localhost only</p><p>Performance Metrics Service 9104 localhost only</p><h3>3. Prerequisites Checklist</h3><p>Confirm every item below before starting. Missing prerequisites are the most common cause of a failed or unstable hub.</p><p>• Oracle Grid Infrastructure installed and healthy on both nodes (crsctl check cluster -all returns online).</p><p>• ASM disk group with free space for an ACFS volume sized for the deployment plus trail-file retention.</p><p>• Identical OS packages, users (oracle, grid), groups, and directory paths on every node.</p><p>• XAG 10.2 or later installer available; it must be installed OUTSIDE the Grid Infrastructure home.</p><p>• Oracle GoldenGate 23ai MA install image for the correct database/OS platform.</p><p>• NGINX available at the version required by your GoldenGate release (23ai on OL8/RHEL8 typically needs NGINX 1.24+ via the appropriate module stream).</p><p>• JRE 8 or higher present on the hub nodes.</p><p>• Network: a free application VIP on the public subnet, plus firewall rules allowing 443 to the VIP and Oracle Net (1521 or your listener port) from the hub to the databases.</p><p>• Source and target databases reachable from the hub, with MAA role-based services planned.</p><p>• The oracle user has sudo permission for the NGINX and VIP operations that require root.</p><p><strong>Important — </strong>Do NOT enable TLS/SSL inside the GoldenGate deployment when using an NGINX reverse proxy — NGINX provides the transport security. Also keep the Performance Metrics Service datastore off the ACFS replication mount, or use the LMDB datastore type, to avoid contention on the shared file system.</p><h3>4. Task 1 — Configure ACFS Shared Storage</h3><p>Create an ASM volume and lay down an ACFS file system that will hold the deployment home and trails. Register it with Clusterware but set autostart to NEVER, because XAG will control when it mounts.</p><h3>4.1 Create the ASM volume and ACFS file system (as the grid user)</h3><p># Create a 60G volume in the DATA disk group</p><p>asmcmd volcreate -G DATA -s 60G ACFS_GG</p><p>asmcmd volinfo -G DATA ACFS_GG # note the Volume Device path</p><p># Make the ACFS file system on the returned device</p><p>/sbin/mkfs -t acfs /dev/asm/acfs_gg-###</p><h3>4.2 Register the file system with Clusterware (as root/grid)</h3><p>srvctl add filesystem \</p><p>-device /dev/asm/acfs_gg-### \</p><p>-path /mnt/acfs_gg \</p><p>-mountowner oracle \</p><p>-user oracle \</p><p>-fstype ACFS \</p><p>-autostart NEVER</p><p>srvctl start filesystem -device /dev/asm/acfs_gg-###</p><p># Confirm the mount and note the resource name for later XAG use</p><p>srvctl status filesystem -device /dev/asm/acfs_gg-###</p><p><strong>Tip — </strong>Record the ACFS Clusterware resource name (for example ora.data.acfs_gg.acfs). You will pass it to XAG with — filesystems so the deployment only starts when its storage is mounted.</p><p>mkdir -p /mnt/acfs_gg/deployments</p><p>chown oracle:oinstall /mnt/acfs_gg/deployments</p><h3>5. Task 2 — Create the Application VIP</h3><p>The application VIP is the stable address clients use. Clusterware moves it to whichever node runs the deployment.</p><p># As root, using the Grid home appvipcfg utility</p><p>$GRID_HOME/bin/appvipcfg create \</p><p>-network=1 \</p><p>-ip=10.10.10.50 \</p><p>-vipname=gghub_vip \</p><p>-user=root</p><p># Allow the oracle user to start/stop the VIP</p><p>$GRID_HOME/bin/crsctl setperm resource gghub_vip -u user:oracle:r-x</p><p># Start and verify</p><p>$GRID_HOME/bin/crsctl start resource gghub_vip</p><p>$GRID_HOME/bin/crsctl status resource gghub_vip</p><p><strong>DNS — </strong>Register gghub-vip in DNS (or /etc/hosts on all clients) so the reverse-proxy URL <a href="https://gghub-vip/">https://gghub-vip:443</a> resolves everywhere.</p><h3>6. Task 3 — Install GoldenGate MA Software on Every Node</h3><p>Install the GoldenGate software into a LOCAL home on each node using the same path. The software is local; only the deployment (Task 4) lives on ACFS.</p><h3>6.1 Interactive install</h3><p>unzip -q V######-01.zip -d /stage/ogg</p><p>cd /stage/ogg/fbo_ggs_Linux_x64_services_shiphome/Disk1</p><p>./runInstaller</p><p># Software Location: /u01/app/oracle/goldengate/gg23ai</p><h3>6.2 Silent install (repeat identically on each node)</h3><p>./runInstaller -silent -nowait \</p><p>INSTALL_OPTION=ORA23c \</p><p>SOFTWARE_LOCATION=/u01/app/oracle/goldengate/gg23ai</p><p><strong>Consistency — </strong>The software home path MUST be identical on all nodes. XAG and relocation rely on matching paths.</p><h3>7. Task 4 — Create the Microservices Deployment on ACFS</h3><p>Run the GoldenGate Configuration Assistant (oggca) ONCE, from the node that currently owns the ACFS mount, and place the deployment home on the shared file system.</p><p>export OGG_HOME=/u01/app/oracle/goldengate/gg23ai</p><p>$OGG_HOME/bin/oggca.sh</p><p>In the assistant, set the following:</p><p>• <strong>Deployment name: </strong>gghub1</p><p>• <strong>Service Manager: </strong>create a new Service Manager; bind services to the loopback/localhost address.</p><p>• <strong>Deployment home: </strong>/mnt/acfs_gg/deployments/gghub1 (on ACFS).</p><p>• <strong>Administrator account: </strong>username oggadmin and a strong password — record these; XAG and REST calls need them.</p><p>• <strong>Security: </strong>do NOT enable TLS/SSL (NGINX handles it).</p><p>• <strong>Metrics datastore: </strong>choose LMDB, or if using Berkeley DB place its home OFF the ACFS replication mount.</p><h3>7.1 Silent creation (optional)</h3><p>$OGG_HOME/bin/oggca.sh -silent \</p><p>-responseFile /stage/ogg/gghub1_oggca.rsp</p><p># Response file sets DEPLOYMENT_NAME, ADMINISTRATOR_USER,</p><p># SERVICEMANAGER_DEPLOYMENT_HOME, OGG_DEPLOYMENT_HOME=/mnt/acfs_gg/…</p><p>Start the deployment manually once to confirm it is healthy, then stop it so XAG can take ownership:</p><p>$OGG_HOME/bin/ServiceManager status || true</p><p># Log in to the web UI locally to confirm the services are green,</p><p># then stop the Service Manager before registering with XAG.</p><h3>8. Task 5 — Install and Configure XAG</h3><p>Install XAG on every node in the same path, outside the Grid Infrastructure home, and put its bin directory on PATH BEFORE the GI bin directory.</p><p># As grid, unzip the XAG image and install to all nodes</p><p>cd /stage/xag</p><p>./xagsetup.sh — install \</p><p>— directory /u01/app/grid/xag \</p><p>— all_nodes</p><p># Add XAG to PATH ahead of the GI home (in the oracle user profile)</p><p>export PATH=/u01/app/grid/xag/bin:$PATH</p><p>agctl query releaseversion</p><p><strong>Why path order matters — </strong>Both GI and XAG ship an agctl. Placing the XAG bin directory first ensures you use the standalone-agent version required for Microservices deployments.</p><h3>9. Task 6 — Register the Deployment with XAG</h3><p>Register the deployment as a Clusterware resource so XAG manages its lifecycle and its dependencies on the ACFS mount and VIP. Confirm exact options with agctl add goldengate — help for your version.</p><p>agctl add goldengate gghub1 \</p><p>— gg_home /u01/app/oracle/goldengate/gg23ai \</p><p>— instance_home /mnt/acfs_gg/deployments/gghub1 \</p><p>— nodes ggnode1,ggnode2 \</p><p>— vip_name gghub_vip \</p><p>— filesystems ora.data.acfs_gg.acfs \</p><p>— oracle_home /u01/app/oracle/goldengate/gg23ai \</p><p>— user oracle \</p><p>— group oinstall</p><h3>9.1 Start and verify</h3><p>agctl start goldengate gghub1</p><p>agctl status goldengate gghub1</p><p># Cross-check the Clusterware view</p><p>crsctl status resource -t | grep -i gghub</p><p><strong>Startup order — </strong>At this point XAG guarantees the ACFS mount and VIP come online before the deployment. NGINX is added to this dependency chain in the next task.</p><h3>10. Task 7 — Configure the NGINX Reverse Proxy</h3><p>NGINX gives you one HTTPS entry point (port 443) and TLS termination in front of the localhost-bound microservices. GoldenGate ships a helper that generates the NGINX configuration.</p><h3>10.1 Install NGINX</h3><p># Oracle Linux 8/9 example (enable the required module stream first)</p><p>sudo dnf module reset nginx -y</p><p>sudo dnf module enable nginx:1.24 -y</p><p>sudo dnf install nginx -y</p><h3>10.2 Generate the proxy configuration</h3><p>export OGG_HOME=/u01/app/oracle/goldengate/gg23ai</p><p>$OGG_HOME/lib/utl/reverseproxy/ReverseProxySettings \</p><p>— user oggadmin \</p><p>— port 443 \</p><p>— host gghub-vip \</p><p>-o /etc/nginx/conf.d/ogg.conf \</p><p><a href="https://localhost:9100/">https://localhost:9100</a></p><p># You will be prompted for the oggadmin password.</p><h3>10.3 Install a certificate, validate, and reload</h3><p># Place server certificate/key referenced by ogg.conf, then:</p><p>sudo nginx -t # validate configuration</p><p>sudo systemctl restart nginx</p><p># Browse to <a href="https://gghub-vip/">https://gghub-vip:443</a> to reach the Service Manager UI.</p><h3>10.4 Make NGINX cluster-managed and add it to the XAG dependency</h3><p>Create an NGINX Clusterware resource so it fails over with the deployment, then modify the XAG resource so NGINX starts before GoldenGate.</p><p># Determine the current filesystems parameter</p><p>agctl config goldengate gghub1 | grep -i “File System”</p><p># Add nginx to the dependency list (as root)</p><p>/u01/app/grid/xag/bin/agctl modify goldengate gghub1 \</p><p>— filesystems ora.data.acfs_gg.acfs,nginx</p><p><strong>Harden access — </strong>After the proxy works, restrict the microservices so they only accept connections from localhost; all external access then flows through the NGINX HTTPS port. Verify direct non-proxy access is refused.</p><h3>11. Task 8 — Prepare Source and Target Databases (MAA)</h3><p>For a hub to survive database failover, Extract and Replicat must connect through role-based services that follow the primary. Perform the database-side setup on both source and target.</p><h3>11.1 Enable logging and create the GoldenGate user</h3><p>— As a DBA on each database (CDB/PDB as appropriate)</p><p>ALTER DATABASE FORCE LOGGING;</p><p>ALTER DATABASE ADD SUPPLEMENTAL LOG DATA;</p><p>ALTER SYSTEM SET ENABLE_GOLDENGATE_REPLICATION=TRUE SCOPE=BOTH;</p><p>CREATE USER c##ggadmin IDENTIFIED BY “****” CONTAINER=ALL;</p><p>GRANT DBA TO c##ggadmin CONTAINER=ALL;</p><p>EXEC DBMS_GOLDENGATE_AUTH.GRANT_ADMIN_PRIVILEGE(‘c##ggadmin’,CONTAINER=&gt;’ALL’);</p><h3>11.2 Create role-based services (they run only on the primary)</h3><p># Source database service</p><p>srvctl add service -db srcdb -service gg_src_svc \</p><p>-preferred srcdb1 -available srcdb2 -role PRIMARY</p><p>srvctl start service -db srcdb -service gg_src_svc</p><p># Target database service</p><p>srvctl add service -db tgtdb -service gg_tgt_svc \</p><p>-preferred tgtdb1 -available tgtdb2 -role PRIMARY</p><p>srvctl start service -db tgtdb -service gg_tgt_svc</p><h3>11.3 Add TNS entries on the hub pointing at those services</h3><p># $OGG_HOME/network/admin/tnsnames.ora (or deployment tns admin)</p><p>GG_SRC = (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=src-scan)(PORT=1521))</p><p>(CONNECT_DATA=(SERVICE_NAME=gg_src_svc)))</p><p>GG_TGT = (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=tgt-scan)(PORT=1521))</p><p>(CONNECT_DATA=(SERVICE_NAME=gg_tgt_svc)))</p><p><strong>Why role-based — </strong>Because the service is defined with -role PRIMARY, after a Data Guard switchover the service starts automatically on the new primary. Extract and Replicat reconnect through the same TNS alias with no reconfiguration.</p><h3>12. Task 9 — Configure Replication (Extract, Path, Replicat)</h3><p>Use the Admin Client (or the web UI) against the hub through the reverse proxy. Store database passwords in the credential store rather than in parameter files.</p><h3>12.1 Connect and set up credentials</h3><p>$OGG_HOME/bin/adminclient</p><p>OGG&gt; connect <a href="https://gghub-vip/services/gghub1">https://gghub-vip:443/services/gghub1</a> as oggadmin</p><p>OGG&gt; add credentialstore</p><p>OGG&gt; alter credentialstore add user c##ggadmin@GG_SRC alias src</p><p>OGG&gt; alter credentialstore add user c##ggadmin@GG_TGT alias tgt</p><h3>12.2 Create the Integrated Extract (source)</h3><p>OGG&gt; dblogin useridalias src</p><p>OGG&gt; add schematrandata SALES allcols</p><p>OGG&gt; register extract ext1 database container (salespdb)</p><p>OGG&gt; add extract ext1 integrated tranlog begin now</p><p>OGG&gt; add exttrail ea, extract ext1</p><p>OGG&gt; start extract ext1</p><h3>12.3 Create a Distribution Path (hub to hub / to receiver)</h3><p>In the Distribution Service, create a path that reads the local trail and sends it to the Receiver Service using the wss (secure WebSocket) target through the reverse proxy. This can be done in the web UI or via REST/Admin Client.</p><p>OGG&gt; add distributionpath path1 \</p><p>source trail ea \</p><p>target wss://gghub-vip:443/services/gghub1/receiver</p><h3>12.4 Create the Integrated Replicat (target)</h3><p>OGG&gt; dblogin useridalias tgt</p><p>OGG&gt; add replicat rep1 integrated exttrail ea</p><p>OGG&gt; start replicat rep1</p><p>OGG&gt; info all</p><h3>13. Task 10 — Validate High Availability</h3><p>Do not consider the build complete until you have proven relocation and failover behave correctly.</p><h3>13.1 Planned relocation (maintenance test)</h3><p>agctl status goldengate gghub1</p><p>agctl relocate goldengate gghub1 — node ggnode2</p><p>agctl status goldengate gghub1</p><p># Confirm the whole stack moved with it:</p><p>crsctl status resource gghub_vip</p><p>srvctl status filesystem -device /dev/asm/acfs_gg-###</p><p>systemctl status nginx</p><h3>13.2 Checks after relocation</h3><p>• The application VIP now resolves to ggnode2 and <a href="https://gghub-vip/">https://gghub-vip:443</a> still loads the UI.</p><p>• Extract and Replicat show RUNNING with no gap in the report files (info all).</p><p>• Trail sequence and checkpoints are continuous — confirming ACFS preserved state.</p><h3>13.3 Unplanned failover test</h3><p>1. On the active node, simulate failure (for example, reboot the node or stop CRS).</p><p>2. Watch Clusterware relocate the VIP, ACFS, NGINX, and deployment to the surviving node.</p><p>3. Confirm replication resumes automatically from the last checkpoint.</p><h3>13.4 Database switchover test</h3><p>1. Perform a Data Guard switchover on the source (and/or target).</p><p>2. Confirm the role-based service restarts on the new primary and Extract/Replicat reconnect via the same TNS alias without manual changes.</p><h3>14. Monitoring, Maintenance, and Patching</h3><p>• <strong>Health at a glance: </strong>agctl status goldengate gghub1 and crsctl status resource -t.</p><p>• <strong>Process lag and status: </strong>use the Performance Metrics Service, the web UI, or info all / stats in the Admin Client.</p><p>• <strong>Enterprise monitoring: </strong>integrate with Oracle Enterprise Manager Cloud Control or scrape the metrics endpoint.</p><p>• <strong>Data consistency: </strong>periodically validate source/target with Oracle GoldenGate Veridata.</p><p>• <strong>Rolling patching: </strong>relocate the deployment to node B, patch node A’s software home, relocate back, then patch node B — keeping replication running throughout.</p><p>• <strong>Backups: </strong>include the ACFS deployment home (parameter files, credential store, trails as needed) in your backup regime.</p><h3>15. Troubleshooting Quick Reference</h3><p><strong>Symptom</strong></p><p><strong>Likely cause</strong></p><p><strong>Action</strong></p><p>Deployment won’t start under XAG</p><p>ACFS not mounted / wrong resource name</p><p>Check — filesystems value; start the ACFS resource; review agctl config</p><p>UI unreachable on 443</p><p>NGINX down or cert issue</p><p>nginx -t; systemctl status nginx; confirm ogg.conf host/port</p><p>VIP not moving on failover</p><p>Permissions or network number</p><p>crsctl setperm on VIP; verify appvipcfg network id</p><p>Extract abends after DB switchover</p><p>Service not role-based</p><p>Recreate service with -role PRIMARY; confirm TNS alias uses the service</p><p>Two agctl versions conflict</p><p>GI agctl on PATH first</p><p>Put /u01/app/grid/xag/bin ahead of the GI bin in PATH</p><p>Metrics/ACFS contention</p><p>Datastore on ACFS mount</p><p>Switch datastore to LMDB or move BDB home off ACFS</p><h3>16. Appendix A — Common agctl Commands</h3><p>agctl add goldengate &lt;name&gt; … # register the deployment</p><p>agctl config goldengate &lt;name&gt; # show configuration &amp; dependencies</p><p>agctl modify goldengate &lt;name&gt; … # change VIP / filesystems / nodes</p><p>agctl start goldengate &lt;name&gt; # start on current/preferred node</p><p>agctl stop goldengate &lt;name&gt; # stop the deployment stack</p><p>agctl status goldengate &lt;name&gt; # where it runs + state</p><p>agctl relocate goldengate &lt;name&gt; — node &lt;node&gt; # planned move</p><p>agctl remove goldengate &lt;name&gt; # deregister</p><h3>17. Appendix B — Build Order Summary</h3><p>1. ACFS shared storage (autostart NEVER).</p><p>2. Application VIP.</p><p>3. GoldenGate software on every node (identical local path).</p><p>4. Deployment created once on ACFS (no TLS; LMDB metrics).</p><p>5. XAG installed outside GI home, PATH ahead of GI.</p><p>6. Deployment registered with XAG (VIP + ACFS dependencies).</p><p>7. NGINX reverse proxy generated, secured, added to XAG dependency.</p><p>8. Databases prepared with role-based services and TNS.</p><p>9. Extract, Distribution Path, and Replicat configured.</p><p>10. Relocation, failover, and switchover tested.</p><p><em>End of document — adapt all names, paths, ports, and sizes to your environment and validate commands against your installed</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dd174833dc53" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>