1. Why ?
2. What is it ?
3. How ?

1. Why ?

The main idea of the 'grid' is to share resources, to make better use of storage, CPU and memory. If a database administrator wishes to implement this idea, he or she must consolidate many databases to one database. One of the concerns of running many applications together in one database is: ‚what will happen, if one of the applications must be restored because of a human error?‘ Tablespace point in time recovery can be used for this purpose, but there are a few prerequisites. Most importantly the tablespaces are strictly separated for each application. Another reason for creating separated databases is security: each customer has his own database. Therefore, there is often a proliferation of smaller databases. Each of them must be maintained, upgraded, each allocates virtual memory and runs background processes thereby wasting resources. Oracle 12c offers another possibility for virtualization, providing isolation at the database level: the multitenant container database holding pluggable databases.

2. What ?

Pluggable databases are logical units inside a multitenant container database, which consists of one multitenant container database and up to 252 pluggable databases. The SGA is shared as are the background processes.

The multitenant container database holds metadata information common for pluggable databases inside the System and the Sysaux tablespace, and there is just one Undo tablespace. The pluggable databases have smaller System and Sysaux tablespaces, containing just their 'personal' metadata. New data dictionary views will make the information available either on pdb (dba_views) or container level (cdb_views).

There are local users, which are known in specific pluggable databases and common users known in all containers.
Pluggable databases can be easily plugged to another multitenant container database and converted from a non-CDB. They can undergo point in time recovery.

3. How ?

Creating a multitenant container database can be done using the database configuration assistant: There you find the new option: Create as Container Database. If you prefer ‚hand made‘ databases you can execute the command from a instance in nomount state:

CREATE DATABASE cdb1 ENABLE PLUGGABLE DATABASE ….

And of course this can also be achieved through Enterprise Manager Cloud.

A freshly created multitenant container database consists of two containers: the root container as the 'rack' and a seed container, a template for future pluggable databases.

There are 4 ways to create other pluggable databases:

1. Create an empty pdb from seed

2. Plug in a non-CDB

3. Move a pdb from another pdb

4. Copy a pdb from another pdb

We will discuss option2: how to plug in a non_CDB into a multitenant container database. Three different methods are available :

1. Create an empty pdb and use Datapump in traditional export/import mode or with
Transportable Tablespace or Database mode. This method is suitable for pre 12c databases.

2. Create an empty pdb and use GoldenGate replication.
When the pdb catches up with the non-CDB, you fail over to the pdb.

3. Databases of Version 12c or higher can be plugged in with the help of the new dbms_pdb Package.

This is a demonstration for method 3:

Step1: Connect to the non-CDB to be plugged in and create an xml File with description of the database.
The xml file is written to $ORACLE_HOME/dbs per default and contains mainly information about the datafiles.

Step 2: Check if the non-CDB is pluggable in the multitenant container database:

Step 3: Create the pluggable database, connected to the Multitenant container database.

With nocopy option the files will be reused, but the tempfile is created anew:
A service is created and registered automatically with the listener:

Step 4: Delete unnecessary metadata from PDB SYSTEM tablespace:
To connect to newly created pdb, edit tnsnames.ora and add entry for new pdb.
Connect to plugged-in non_CDB and clean up Data Dictionary to remove entries now maintained in multitenant container database. As all kept objects have to be recompiled it will take a few minutes.

Step 5: The plugged-in database will be automatically synchronised by creating common users and roles when opened the first time in read write mode.

Step 6: Verify tablespaces and users:

There is only one local tablespace (users) and one local user (scott) in the plugged-in non_CDB pdb_orcl.
This method of creating plugged_in non_CDB from is fast and easy for 12c databases.

The method for deplugging a pluggable database from a CDB is to create a new non_CDB and use the the new full transportable feature of Datapump and drop the pluggable database.

About the Author:

Gerlinde has been working for Oracle University Germany as one of our Principal Instructors for over 14 years. She started with Oracle 7 and became an Oracle Certified Master for Oracle 10g and 11c. She is a specialist in Database Core Technologies, with profound knowledge in Backup & Recovery, Performance Tuning for DBAs and Application Developers, Datawarehouse Administration, Data Guard and Real Application Clusters.