Bulletproof EBS on OCI (Episode 2/4)

Because we utilized the Autonomous Recovery Service, the database restore is largely driven by the OCI Console, but there are specific parameters required to ensure EBS connectivity later.

The Workflow:

We initiated a "Restore to new DB System" from our locked backup window.

"Create database" from a backup

Critical Configuration Details:

• Grid Infrastructure: We explicitly selected "Oracle Grid Infrastructure" (required for our architecture), rather than the Logical Volume Manager.

• Hostname & DB Name: To minimize reconfiguration effort, we matched the source hostname (master240125db) and Database Name (CDBMAST).

• Shape: We selected an appropriate shape (Standard E5) to balance cost vs. restore speed.

Result: The 4-OCPU system restored fully in approx. 1.75 hours. While this ran, we moved immediately to the Application Tier.

Phase 2: The App Tier Restore (The Manual Lift)

Since we used our "Workaround Strategy" (tarballed backups in locked Object Storage) for the App Tier, this required a manual, validatable procedure.

1. OS Preparation

We spun up a vanilla Oracle Linux 8 compute instance in our recovery subnet and prepped the prerequisites:

• Installed python36-oci-cli (to talk to our bucket).

• Installed EBS pre-reqs (oracle-ebs-server-R12-preinstall).

• The "Gotcha": We applied a specific fix for libXm.so to ensure Oracle Forms would launch correctly later:

cd /usr/lib && sudo ln -s libXm.so.4.0.4 libXm.so.2

2. The Retrieval

This was the moment of truth. Could we read from the locked bucket? We used Instance Principals for authentication to stream and extract the backup on the fly:

oci os object list --auth instance_principal --bucket-name Master_Immutable_Backup | grep tgz
# choose the latest:
oci os object get --bucket-name Master_Immutable_Backup --auth instance_principal --name bkp_fullMAST_master240125app01_20251119_012240.tgz --file - | sudo tar -zxp

Result: The restore of the /u01 volume completed in roughly 25 minutes – finishing well before the database was ready.

In this case we restored the (very simple, single-node) instance to local Block Storage. In real scenarios you'll probably want to restore to File Storage Service (FSS) and have multiple apps-tier nodes.

Phase 3: The Handshake (Reconfiguration)

Once both tiers were online, we had to stitch them together. Since this was a restore to a new (test-restore-) subnet, networking and configuration files needed updates. We put a manual entry to /etc/hosts so that the apps- and db-tier can talk to each other; if you restore to a subnet with the same domain than the source (in real desaster) this is not necessary though.

Since restoring the database provides a "new" Oracle Home (not a backup of the source), we have to eventually apply individual patches and especially recreate the appsutil directory. Then we run the specific EBS-tool txkPostPDBCreationTasks.pl to register the restored PDB with the system.

##for proper utl_file-setup:
sudo mkdir -p /u01/app/oracle/product/19.0.0.0/temp/MAST
sudo chown -R oracle:oinstall /u01/app/oracle/product/19.0.0.0/temp/

cd $ORACLE_HOME
unzip /tmp/appsutil.zip # copy from /u01/install/APPS/fs?/inst/apps/MAST_mastebsapp01/admin/out/appsutil.zip on the Apps-Tier

cd $ORACLE_HOME/appsutil
cp -r $ORACLE_HOME/jdk/jre .
cp $ORACLE_HOME/jlib/orai18n.jar $ORACLE_HOME/appsutil/jre/lib/ext

cd $ORACLE_HOME/appsutil
. ./txkSetCfgCDB.env dboraclehome=/u01/app/oracle/product/19.0.0.0/dbhome_1
export ORACLE_SID=CDBMAST
export TNS_ADMIN=/u01/app/oracle/product/19.0.0.0/dbhome_1/network/admin

sqlplus system
# must work without errors and also no warnings (password to expire)

perl $ORACLE_HOME/appsutil/bin/txkPostPDBCreationTasks.pl \
-dboraclehome=$ORACLE_HOME \
-outdir=/u01/app/oracle/product/19.0.0.0/dbhome_1/appsutil/log \
-cdbname=$ORACLE_SID \
-dbuniquename=$ORACLE_UNQNAME \
-cdbsid=$ORACLE_SID \
-pdbsid=MAST \
-appsuser=apps \
-israc=no \
-virtualhostname=master240125db.restoretestsubn.ebsnetwork.oraclevcn.com \
-dbport=1521 \
-generateutlfiledir=yes \
-servicetype=dbsystem

##Important Post-Steps:
sqlplus / as sysdba
alter pluggable database all save state;

srvctl stop database -d $ORACLE_UNQNAME
cd $ORACLE_HOME/bin ; chopt enable olap
cd $ORACLE_HOME/bin ; chopt enable partitioning

srvctl stop listener
srvctl start listener
srvctl start database -d $ORACLE_UNQNAME

After updating the /etc/hosts file to point to the new DB IP, we ran the standard EBS configuration tools. We also verified the utl_file_dir paths were correctly registered in the database context.

. setenv_run.sh
sqlplus apps
# should work properly
adautocfg.sh
# should have no errors
. setenv_patch.sh
adautocfg.sh
# should only complain about EBS_LOGON Trigger
# In a new session (make sure that ssh to localhost works with the default priv-key):
. setenv_run.sh
./myadop.sh -validate

A modern EBS environment isn't just Forms and Reports. We also reconfigured:

• ORDS: Re-installed/configured to point to the new DB service for APEX support.

• Enterprise Command Center (ECC): Created a new DBC file and updated some IP Addresses in the config files.

Phase 4: Startup, Validation & Cleanup

With those preparations in place we can finally startup the E-Business Suite apps-tier.

adstrtal.sh

In order to access the environment we created a restoretest-loadbalancer pointing to the Apps- and ORDS Backends in the restore-subnet.

We pointed our local browser hosts file to the Restore Load Balancer (eventually locally overriding the usual hostname with the restore-loadbalancer-IP) and ran the victory lap:

1. Login: Successful via HTTPS.

2. Forms: Opened successfully (thanks to the library fix!).

3. Concurrent Managers: Active and processing.

4. ECC: Works.

5. APEX: Accessible.

Once validated, we terminated the Compute instances and the Database System to keep the infrastructure cost of this entire validation drill negligible.

This project at Broadpin demonstrated that you can achieve immutable, ransomware-proof backups for Oracle E-Business Suite on OCI without sacrificing recoverability.

By combining the Autonomous Recovery Service with a creative Object Storage strategy, we ensured the customer is protected against the worst-case scenario.