Technical Blog

Technical and Product News and Insights from Rackspace

Use the Oracle Database refreshable clone featureā€”Part One: Introduction

The OracleĀ® Database 18c for Enterprise Edition on Cloud bundles engineered systems (Exadata) with many new and improved features. Release 18c improves the refreshable clone feature from the 12c release 2 by adding a quick switchover or failover facility. You can also create a snapshot carousel to clone or replicate pluggable databases (PDBs) in a multitenant environment.

For production environments, use a snapshot carousel with the refreshable clone PDB feature to make database administration (DBA) activities easy, such as safeguarding against an unforeseen outage or troubleshooting logical corruption issues. This post describes the practical use of refreshable clones with manual and automatic switchover cases. You can use it with the snapshot carousel feature to identify and handle the following database health problems:

  • Production database unavailable

  • Performance issues

  • Logical corruption in the Oracle database

This post is part one of a two-part blog post series. Read on to learn about the refreshable clone PDB, how it works, and when to use it.

The second part of the series describes how to create, configure, maintain, and drop a refreshable clone PDB.

Cloning and snapshot questions

This section provides answers to some basic questions about clone and snapshot carousels.

What is cloning?

When you clone an instance, you take a backup of it and restore the backup elsewhere. Typically, you restore it on another machine with the same directory structure. You can also restore it on the same machine by changing the Oracle System ID (SID) and the database name. You can use a clone of a production instance on a test machine to try out what-if scenarios, such as changing the init.ora parameter or altering your code, and so on.

How does PDB cloning work?

You can use PDB cloning to clone a PDB in a multitenant environment. For local or remote PDBs, use a refreshable clone or snapshot carousel to create PDB clones in a local container database.

For remote PDB cloning, consider the following:

  • To use a refreshable clone, you need to have a database link for the PDB and disable the cloned database.

  • To use a snapshot carousel, you need to create a typical PDB clone by using a snapshot. Then, use the database links or unplug and plug into another container database. You can also create a refreshable clone from a cloned PDB.

What is a PDB snapshot?

A PDB snapshot is a point-in-time copy of a PDB. Create snapshots manually by using the SNAPSHOT clause of CREATE PLUGGABLE DATABASE (or ALTER PLUGGABLE DATABASE) command, or automatically by using the EVERY interval clause.

A PDB snapshot carousel maintains a library of recent PDB snapshots that you can use to perform a point-in-time recovery and to clone a PDB.

Refreshable clone PDB

You can clone the PDB by using the refreshable clone feature, which protects the database from data corruption and disasters with minimal data loss, depending on the refresh interval and redo generation rate. You can use the refreshable clone database as a replica on which to resume certain low-load, non-critical applications. You can set the refreshable clone database to update automatically at set intervals or manually by using redo log application.

Figure 1 shows the architecture of refreshable clone processing. It shows the main components and processes and illustrates the relationship between the production database and the refreshable clone database. This diagram shows cloning the pluggable database PDB1 of container database CDB1 to another container database CDB2. This action results in a hot-cloned version of PDB1 named PDB1_REF_CLONE.

Figure 1 Figure 1

Figure 1

Refresh mode options

You can change refresh modes by setting up an environment with the following modes:



  • NONE

Create and work with a refreshable clone

Use the following statement to clone the source PDB and configure the clone to be refreshable. Refreshing the clone PDB updates it with redo data accumulated since the last redo log apply.


Use the following statement to change the current mode of a refreshed or disabled refreshable clone and convert it to a fully functional PDB:


If you don’t clone PDBs frequently to avoid performance degradation, the clone data gets stale. A refreshable clone PDB solves this problem. When a refreshable clone gets stale, you can quickly refresh it with a recent redo.

Typically, you maintain a master refreshable clone of a production PDB and then take snapshot clones of the master for development and testing.

Use the following statement to reverse the roles for source and clone PDBs:


You can simplify this switchover process, as shown in the following figures:

Figure 2 Figure 2

Figure 2

Figure 3 Figure 3

Figure 3

This switchover capability is useful in the following situations.

Planned switchover of a refreshable clone

In Figure 3, CDB1, which hosts the source PDB, PDB1, might experience significantly more overhead than CDB2, which hosts the clone PDB, PDB1_REF_CLONE. To achieve better load balancing, you can reverse the roles of PDBs by converting the clone to the new source PDB and the source PDB into the new clone.

Execute this role transition by using the following command on the current primary database:


After this command completes, PDB1_REF_CLONE in CDB2 assumes the primary role. CDB1 now maintains the replica. All connections to production connect to the new primary, which is now CDB2. You lose no more than two minutes of transactions, assuming that the refreshes kept up with the redo generation rate from the source.

Unplanned switchover of a refreshable clone

If the source PDB has an unplanned failure, you can switch the clone PDB to the new source PDB and resume the normal operations.

Be sure to test your environment by using realistic transaction volumes to ensure that the process of refreshing the replica can keep up with the redo generation rate.

How do refreshable clones differ from Data Guard?

Oracle introduced the high availability feature with Data Guard and standby databases. The following elements distinguish refreshable clones from Data Guard:

  • Data Guard provides high availability to protect the database from disasters and data corruption in real-time provided switchover and failover to standby databases. You can also use a Data Guard standby database for load sharing by using the refreshable clone PDB feature. Data Guard works at a CDB level, and you can’t have a switchover or failover at the PDB level.

  • Because of the lag between the initiation and completion of the switchover, Data Guard is more effective than just maintaining a refreshable clone. During this lag, transactions to the primary database might not be applied to or synced with the read-only database before you switch the roles. As a result, you might lose those transactions.

  • Data Guard has a maximum limit of 30 standby databases, but you can have as many refreshable clones as you need.

To strengthen the refreshable clone PDB feature for high availability and have almost no data loss, set the REMOTE_RECOVERY_FILE_DEST parameter to archive the source PDB’s log location.


You should not consider the refreshable clone PDB feature as a replacement for Data Guard from the perspective of high availability. However, you can use a refreshable clone to maintain a replica database on another server.

This post described how to use refreshable PDBs as replicas so that you can resume certain low-load, non-critical applications operations, whether the switchover is a planned or an unplanned event. Keep in mind that you should consider the switchovers from the point of view of Recovery Time Objectives (RTOs, time to resume operations) and Recovery Point Objectives (RPOs, such as achieving minimal data loss).

Part Two of the series demonstrates the refreshable clone feature.

Use the Feedback tab to make any comments or ask questions. You can also chat now to start the conversation.

Learn more about Databases.

post avatar
Tejashkumar Patel

I am a Database Administrator with more than a decade of experience in various RDBMSs, Big Data and NoSQL database technologies, and the cloud platform.

Share this information: