Oracle AI 26ai On-Premise: A New Era for Data Guard and Multi-tenant

Author: Paul Sammy – Oracle ACE, 1st March 2026

This Blog will take you through the step by step setup of a Dataguard Pluggable database (DGPDB) setup.

github repo: psammy-github/blog-sql-scripts: Repo for SQL scripts

*No cartoon characters were harmed in the creation of this image

Oracle’s strategic direction toward Multitenant architecture is no longer a gradual transition — it’s the standard. Pluggable Databases (PDBs) are no longer the “new kid on the block.” In Oracle AI 26AI, they are the only kid on the block.

And that shift has some big implications for Disaster Recovery.

The Historical Limitation

In the Multitenant world — up until now — a Data Guard configuration operated at the Container Database (CDB) level.

That meant:

Data Guard was configured for the entire CDB.
All PDBs within that CDB moved together.
You couldn’t selectively fail over or switch over individual PDBs.
Disaster Recovery was an “all-or-nothing” operation.

If you wanted granularity at the PDB level? That simply wasn’t possible.

What’s Changed in 26ai (Change was introduced in Oracle 21c)

With Oracle AI 26ai, we can now deploy two primary CDBs, and the PDBs within those CDBs can operate independently as primary or standby PDBs.

This means:

A PDB in one CDB can function as a primary.
That same PDB can have its own standby in another CDB.
Other PDBs in the same CDB can operate differently.
Failover and switchover can occur at the PDB level, not just the CDB level.

For anyone running consolidated environments or looking for more granular DR strategies, this is a major shift.

The Lab Setup

Phase 1: Standard Settings

SPFILES, ARCHIVELOG mode, FLASHBACK*, FORCE LOGGING,

This is pretty standard for setting up Dataguard.

Lets run STEP1_DGPDB_precheck_setup.sql on both our CDBs

sqlplus / as sysdba @STEP1_DGPDB_PRE_setup_check.sql

Setup Phase 1 the Archive Destination

Remediating Settings if needed:

Create SPFILE

create spfile from pfile;
-- restart database

Place database in archive log mode

-- in mount mode
startup mount
alter database archivelog;

Enable force logging

alter database force logging;

Optional but recommended turn flashback database on

alter database flashback on;

Enable use of password file for remote SYS login

alter system set remote_login_passwordfile=EXCLUSIVE scope=spfile;
-- restart database

Phase 2: DGPDB Pre Settings

sqlplus / as sysdba

1. Update LOG_ARCHIVE_DEST_1 parameter on both CDBs’

alter system set log_archive_dest_1='LOCATION=USE_DB_RECOVERY_FILE_DEST VALID_FOR=(ALL_LOGFILES,ALL_ROLES)';

2. Update STANDBY_FILE_MANAGEMENT on both CDBs’

alter system set standby_file_management = AUTO scope = both;

3. Update Dataguard Broker parameters on both CDBs’

In ASMCMD create directory structure using mkdir

-- ENGLAND cluster
mkdir +FRA/ENGLAND/DATAGUARDCONFIG
mkdir +DATA/ENGLAND/DATAGUARDCONFIG

-- CANADA cluster
mkdir +FRA/CANADA/DATAGUARDCONFIG
mkdir +DATA/CANADA/DATAGUARDCONFIG

3a. CDB: ENGLAND

ALTER SYSTEM SET DG_BROKER_CONFIG_FILE1='+DATA/ENGLAND/DATAGUARDCONFIG/dr2england.dat' scope=both sid='*';
ALTER SYSTEM SET DG_BROKER_CONFIG_FILE2='+FRA/ENGLAND/DATAGUARDCONFIG/dr1england.dat' scope=both sid='*';

ALTER SYSTEM SET dg_broker_start=TRUE scope=both sid='*';

3b. CDB: CANADA

ALTER SYSTEM SET DG_BROKER_CONFIG_FILE1='+DATA/CANADA/DATAGUARDCONFIG/dr1canada.dat' scope=both sid='*';
ALTER SYSTEM SET DG_BROKER_CONFIG_FILE2='+FRA/CANADA/DATAGUARDCONFIG/dr2canada.dat' scope=both sid='*';

ALTER SYSTEM SET dg_broker_start=TRUE scope=both sid='*';

Phase 3: Update tnsnames.ora on all nodes

Sample tnsnames.ora
We need entries for our CDB and our PDB’s
My PDB’s are RAC so utilise SCAN and failover

ENGLAND= (DESCRIPTION= (ADDRESS= (PROTOCOL= TCP) (HOST= linrac-scan-a) (PORT= 1521))
	   (CONNECT_DATA= (SERVER= DEDICATED) (SERVICE_NAME= ENGLAND)))
	
PROD= (DESCRIPTION= (ADDRESS= (PROTOCOL=TCP) (HOST=linrac-scan-a) (PORT=1521))
        (CONNECT_DATA= (SERVER=DEDICATED) (SERVICE_NAME=ENGLAND_prod) (FAILOVER_MODE= (TYPE=select) (METHOD=basic))))
	
CANADA= (DESCRIPTION= (ADDRESS= (PROTOCOL= TCP) (HOST= linrac-scan-b) (PORT= 1521))
	  (CONNECT_DATA= (SERVER= DEDICATED) (SERVICE_NAME= CANADA)))

TEST= (DESCRIPTION= (ADDRESS= (PROTOCOL=TCP) (HOST=linrac-scan-b) (PORT=1521))
	  (CONNECT_DATA= (SERVER=DEDICATED) (SERVICE_NAME=CANADA_test) (FAILOVER_MODE= (TYPE=select) (METHOD=basic))))

Don’t forget to test those connections
TNS Ping CDB DB from each host

tnsping ENGLAND
tnsping CANADA
tnsping PROD
tnsping TEST

Phase 4: Create Oracle Wallets and store CDB sys passwords

create a common directory on all nodes to store the Oracle Wallet

As user oracle
Environment should be set for CDB using oraenv
*I had intended to create these on an ACFS f/s, but have issue on 26AI/OEL9 but had error creating filesystem
- So I used normal o/s filesystem local to each node and created on each node

. oraenv

mkdir /u01/app/oracle/wallet
chmod 700 /u01/app/oracle/wallet

Create Wallet
- update CDB name that are referenced in tnsnames.ora
- update password to correct SYS password example below uses “Oracle123!”

mkstore -wrl /u01/app/oracle/wallet -createCredential ENGLAND sys "Oracle123!"
mkstore -wrl /u01/app/oracle/wallet -createCredential CANADA sys "Oracle123!"

Check Wallet Contents

mkstore -wrl /u01/app/oracle/wallet -listCredential

Restart broker on all Instances

alter system set dg_broker_start=FALSE;
alter system set dg_brokder_start=TRUE;

Test Wallet is working

sqlplus /@ENGLAND as sysdba

sqlplus /@CANADA as sysdba

Phase 5: Update sqlnet.ora on all nodes

Need to specify directory used for wallet location example below

WALLET_LOCATION = (SOURCE = (METHOD=file) (METHOD_DATA = (DIRECTORY = /u01/app/oracle/wallet)))
SQLNET.WALLET_OVERRIDE = TRUE

Phase 6: Update Dataguard CDB level configuration

On one of the CDBs only create the Dataguard CDB level configuration

dgmgrl /@ENGLAND

create configuration 'ENGLAND' as connect identifier is ENGLAND;
add configuration 'CANADA' connect identifier is CANADA;
enable configuration all;
show configuration;

dgmgrl /@CANADA

show configuration;

Phase 7: Create new DGPDB_INT user in Dataguard

Connect to one of the CDB’s only using dgmgrl
Ensure password is complex enough or will error
enter password for new DGPDB_INT user as an example: BatManLovesPDBs##
- Should be complex to beat password rules
- Use the same password for both CDB’s
Internally each DGPDB_INT user has a dblink created to communicate
- DB_LINK and credential information is stored in data_guard_site$
If needed the command can be rerun to reset password

dgmgrl /@ENGLAND

edit configuration prepare dgpdb;

Phase 8: Update Dataguard PDB level configuration

This is done before we create the actual standby databases
We must specify a mandatory pdbfilenameconvert parameter to pattern match datafiles coming from primary to standby
Primary PDB must be open in read write mode
Run command on the CDB that we are adding the PDB
- Noticed if not done receive “Created process to run a program, but command failed.”
PDB entries are registered in DGPDB_SITE$

7a. ENGLAND CDB

dgmgrl /@ENGLAND

add pluggable database test at ENGLAND source is test at CANADA pdbfilenameconvert is "'/CANADA/TEST/','/ENGLAND/TEST/'";

7b. CANADA CDB

dgmgrl /@CANADA


add pluggable database prod at CANADA source is prod at ENGLAND pdbfilenameconvert is "'/ENGLAND/PROD/','/CANADA/PROD/'";

On Both CDB’s check:

Its ok to see : ORA-16914: Redo Apply Services were not started at the standby pluggable database.
- As we have not started Redo Apply and the Standby PDB does not exit yet

show pluggable database test at ENGLAND;
show pluggable database prod at CANADA;

Phase 9: Copy the primary PDB datafiles to the standby PDB using RMAN

8a. Run RMAN script to copy the PDB datafiles from the primary PDB to standby PDB using RMAN, example below:

rman target /@ENGLAND auxiliary /@CANADA

run {
backup as copy pluggable database prod auxiliary format '+DATA';
}

8c. Rename datafiles

Review RMAN log to see which datafiles are source and target

Rename files based on LOG file example below
Log on to new standby PDB to check filenames

sqlplus /@CANADA as sysdba

Switch to standby PDB

show pdbs

alter session set container=PROD;

query file locations which need to be updated

select name from v$datafile;

Based on files in log where you can match file name to the filename for restored file update the database
- we can see standby DB thinks datafiles are in
  - +DATA/CANADA/PROD/
- we can see from rman log datafiles were restored to
  - +DATA/CANADA/<GUID Hash Value>/DATAFILE/
- Example rename below
  - Run SQL inside the Standby PDB
  - syntax: alter database rename file ‘<incorrect existing location>’ to ‘rman correct restored location’;

set lines 180
select 'alter database rename file '||''''||name||''''||' to '||''''||'REPLACE WITH NEW RMAN RESTORE NAME'||''''||';'
from v$datafile
order by file#
/

Confirm updated locations

select name from v$datafile;

Repeat the same for CDB CANADA

8b. Run RMAN script to copy the PDB datafiles from the primary PDB to standby PDB using RMAN, example below:

note had to place test in quotes as reserved word below ‘TEST’

rman target /@CANADA auxiliary /@ENGLAND

run {
backup as copy pluggable database 'TEST' auxiliary format '+DATA';
}

8c. Rename datafiles

Review RMAN log to see which datafiles are source and target

Rename files based on LOG file example below
Log on to new standby PDB to check filenames

sqlplus /@ENGLAND as sysdba

Switch to standy PDB

show pdbs

alter session set container=TEST;

query file locations which need to be updated

select name from v$datafile;

Based on files in log where you can match file name to the filename for restored file update the database
- we can see standby DB thinks datafiles are in
  - +DATA/ENGLAND/TEST
- we can see from rman log datafiles were restored to
  - +DATA/ENGLAND/<GUID Hash Value>/DATAFILE/
- Example rename below
  - Run SQL inside the Standby PDB
  - syntax: alter database rename file ‘<incorrect existing location>’ to ‘rman correct restored location’;

select 'alter database rename file '||''''||name||''''||' to '||''''||'REPLACE WITH NEW RMAN RESTORE NAME'||''''||';'
from v$datafile
order by file#
/

Confirm updated locations

select name from v$datafile;

Phase 10: Add Standby Redo logs at PDB level

Logon to standby PDB test in CDB ENGLAND
- Standby Redo logs are created on PDB not CDB
Add 1 more log group non multiplexed
1 group per thread
Standby redo logs are visible on PDB but with CON_ID=0

Standby PDB test @ Primary CDB England:

sqlplus /@ENGLAND as sysdba

alter session set container=TEST;

select group#,thread#,bytes from v$standby_log;

ALTER DATABASE ADD STANDBY LOGFILE thread 1;
ALTER DATABASE ADD STANDBY LOGFILE thread 1;
ALTER DATABASE ADD STANDBY LOGFILE thread 1;

ALTER DATABASE ADD STANDBY LOGFILE thread 2;
ALTER DATABASE ADD STANDBY LOGFILE thread 2;
ALTER DATABASE ADD STANDBY LOGFILE thread 2;

select group#,thread#,bytes from v$standby_log;

Logon to standby PDB prod in CDB CANADA
Add 1 more log group non multiplexed
1 group per thread

Standby PDB prod @ Primary CDB Canada:

sqlplus /@CANADA as sysdba


alter session set container=PROD;

select group#,thread#,bytes from v$standby_log;

ALTER DATABASE ADD STANDBY LOGFILE thread 1;
ALTER DATABASE ADD STANDBY LOGFILE thread 1;
ALTER DATABASE ADD STANDBY LOGFILE thread 1;

ALTER DATABASE ADD STANDBY LOGFILE thread 2;
ALTER DATABASE ADD STANDBY LOGFILE thread 2;
ALTER DATABASE ADD STANDBY LOGFILE thread 2;

select group#,thread#,bytes from v$standby_log;

Phase 11: Enable redo apply on Standby PDBs

CDB England

dgmgrl /@ENGLAND <<EOF
edit pluggable database TEST at ENGLAND set state='APPLY-ON';
EOF

sqlplus /@ENGLAND as sysdba <<EOF
alter system archive log current;
alter system archive log current;
EOF

Check and set FAL_SERVER and FAL_CLIENT

CDB CANADA

dgmgrl /@CANADA <<EOF
edit pluggable database PROD at CANADA set state='APPLY-ON';
EOF


sqlplus /@CANADA as sysdba <<EOF
alter system archive log current;
alter system archive log current;
EOF

set FAL_SERVER and FAL_CLIENT, for some reason these were not set by DGBroker

sqlplus /@ENGLAND as sysdba <<EOF
alter system set FAL_SERVER=CANADA scope=BOTH;
alter system set FAL_CLIENT=ENGLAND scope=BOTH;
EOF

sqlplus /@CANADA as sysdba <<EOF
alter system set FAL_SERVER=ENGLAND scope=BOTH;
alter system set FAL_CLIENT=CANADA  scope=BOTH;
EOF

Phase 12: Health check setup

CDB ENGLAND

Run parameter post check

sqlplus /@ENGLAND as sysdba @STEP2_DGPDB_POST_setup_check.sql

Check Status in dgmgrl

dgmgrl /@ENGLAND <<EOF
show configuration verbose;
show database ENGLAND;
show pluggable database test at ENGLAND;
validate pluggable database test at ENGLAND;
EOF

Check logs
- Alertlog

tail -20 $ORACLE_BASE/diag/rdbms/*/$ORACLE_SID/trace/alert_${ORACLE_SID}.log

Dataguard Log

tail -20 $ORACLE_BASE/diag/rdbms/*/$ORACLE_SID/trace/drc${ORACLE_SID}*.log

select * from gv$dataguard_process where con_id in (select con_id from v$pdbs where name in ('PROD','TEST'));

select * from gv$dataguard_stats;

select * from DGPDB_SITE$;
select * from data_guard_site$;

CDB CANADA

Run parameter post check

sqlplus /@CANADA as sysdba @STEP2_DGPDB_POST_setup_check.sql

Check Status in dgmgrl

dgmgrl /@CANADA <<EOF
show configuration verbose;
show database CANADA;
show pluggable database prod at CANADA;
validate pluggable database prod at CANADA;
EOF

Check logs
- Alertlog

tail -20 $ORACLE_BASE/diag/rdbms/*/$ORACLE_SID/trace/alert_${ORACLE_SID}.log

Dataguard Log

tail -20 $ORACLE_BASE/diag/rdbms/*/$ORACLE_SID/trace/drc${ORACLE_SID}*.log

select * from gv$dataguard_process where con_id in (select con_id from v$pdbs where name in ('PROD','TEST'));

select * from gv$dataguard_stats;

select * from DGPDB_SITE$;
select * from data_guard_site$;

Phase 13: Perform switchover of PDBs’

Switchover from CDB ENGLAND PDB PROD to CDB CANADA PDB PROD

dgmgrl /@CANADA

show pluggable database prod at CANADA;
VALIDATE PLUGGABLE DATABASE prod at CANADA;
SWITCHOVER TO PLUGGABLE DATABASE PROD at CANADA;

Switchover from CDB CANDA PDB TEST to CDB ENGLAND PDB TEST

Clusterware did not seem to open new PDB so open it and optionally save state

sqlplus /@CANADA as sysdba <<EOF
alter pluggable database prod open;
alter pluggable database prod save state; 
EOF

dgmgrl /@ENGLAND

show pluggable database test at ENGLAND;
VALIDATE PLUGGABLE DATABASE test at ENGLAND;
SWITCHOVER TO PLUGGABLE DATABASE TEST at ENGLAND;

Clusterware did not seem to open new PDB so open it and optionally save state

sqlplus /@ENGLAND as sysdba <<EOF
alter pluggable database test open;
alter pluggable database test save state; 
EOF

DGPDB 26AI Limitations

DGPDB only supports MAX PERFORMANCE mode
No more than 2 CDB’s for each configuration
No cascading standby PDB’s
Entire redo stream is shipped for whole CDB not just targeted at PDB level
Does not support automatic block repair
Can’t reinstate a failover with Flashback
Does not support Fast Start failover
Does not support Application Continuity
Does not support Rolling Maintenance

DGPDB 26AI Features

Supports Real Time Query
Supports DML redirect

Remove Dataguard Configuration and Standby Databases

WARNING this is destructive PDBs’ are removed

1a) on CDB ENGLAND

Swich PDBs’ back to their original CDB

dgmgrl /@ENGLAND

VALIDATE PLUGGABLE DATABASE prod at ENGLAND;
SWITCHOVER TO PLUGGABLE DATABASE PROD at ENGLAND;

VALIDATE PLUGGABLE DATABASE test at CANADA;
SWITCHOVER TO PLUGGABLE DATABASE TEST at CANADA;

Stop Apply process
- Catering for all possibilities

dgmgrl /@ENGLAND <<EOF
edit pluggable database TEST at ENGLAND set state='APPLY-OFF';
edit pluggable database TEST at CANADA set state='APPLY-OFF';
edit pluggable database PROD at ENGLAND set state='APPLY-OFF';
edit pluggable database PROD at CANADA set state='APPLY-OFF';
EOF

Remove Standby PDB and delete datafiles

dgmgrl /@ENGLAND

remove pluggable database test at ENGLAND REMOVE DATAFILES;
remove pluggable database prod at CANADA REMOVE DATAFILES;

disable configuration ENGLAND;
disable configuration CANADA;


-- must be in correct order
remove configuration CANADA;
remove configuration ENGLAND;

-- show configuration show indicate no configuration exists
show configuration;

1b) On CDB CANADA

dgmgrl /@CANADA


remove configuration CANADA;
show configuration;

Revert SPFILE changes on both CDBs, example below:
Bounce Database following changes

sqlplus /@ENGLAND as sysdba

sqlplus /@CANADA as sysdba


alter system reset LOG_ARCHIVE_CONFIG scope=spfile sid='*';
alter system reset LOG_ARCHIVE_DEST_1 scope=spfile sid='*';
alter system reset LOG_ARCHIVE_DEST_STATE_2 scope=spfile sid='*';
alter system reset DG_BROKER_START scope=spfile sid='*';
alter system reset DG_BROKER_CONFIG_FILE1 scope=spfile sid='*';
alter system reset DG_BROKER_CONFIG_FILE2 scope=spfile sid='*';
alter system reset STANDBY_FILE_MANAGEMENT scope=spfile sid='*';
alter system reset FAL_SERVER scope=spfile sid='*';
alter system reset FAL_CLIENT scope=spfile sid='*';

Bounce both databases

srvctl stop database -db ENGLAND
srvctl start database -db ENGLAND

srvctl stop database -db CANADA
srvctl start database -db CANADA

References

https://mylearn.oracle.com/ou/course/oracle-ai-database-data-guard-administration

Set up an Oracle Data Guard at the Pluggable Database Level

Oracle Data Guard per Pluggable Database is available today! | maa

dbi Blog

Oracle Data guard per-PDB (DGPDB) in 23ai – DatabaseVerse: Journey into the World of Databases

DG PDB : Oracle Data Guard per Pluggable Database in Oracle Database 21c (21.7 Onward) – ORACLE-BASE

Set up an Oracle Data Guard at the Pluggable Database Level

Cookie	Duration	Description
cookielawinfo-checkbox-analytics	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Analytics".
cookielawinfo-checkbox-functional	11 months	The cookie is set by GDPR cookie consent to record the user consent for the cookies in the category "Functional".
cookielawinfo-checkbox-necessary	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookies is used to store the user consent for the cookies in the category "Necessary".
cookielawinfo-checkbox-others	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Other.
cookielawinfo-checkbox-performance	11 months	This cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Performance".
viewed_cookie_policy	11 months	The cookie is set by the GDPR Cookie Consent plugin and is used to store whether or not user has consented to the use of cookies. It does not store any personal data.