Oracle Real Application Clusters 12c, Release 2 enhancement for Oracle Data Guard

Introduction

Oracle Data Guard feature enables customers to create one or more synchronized copies of their Database at a standby site to protect against disasters at the Primary site. Customers typically deploy Oracle Data Guard in combination with Oracle Real Application Clusters (RAC) to avail of the Scalability and Availability benefits that Oracle RAC provides while ensuring that there is no Data loss due to site failures.

In a typical configuration, an Oracle RAC Primary Database is configured to ship redo to one or more Oracle RAC standby sites. Prior to Oracle 12c Release 2, only one instance of Oracle RAC standby would apply Redo.  As shown in the image below, the RFS (“Remote File server”) process on each Standby RAC instance writes to the SRL “Standby Reo Log” respectively however, the MRP ("Managed recovery process") is only running on ONE Instance .

New Feature introduced in Oracle RAC 12c, Release 2

Starting with Oracle RAC 12c Release 2, each instance on the Standby Site can participate in Redo application. This enhancement improves the performance of Redo apply as all the Oracle RAC instances at the standby site can now apply apply redo especially. Environments where the Redo Apply is CPU bound would benefit greatly with this feature. The ability to apply Redo from multiple instances will work with both Data Guard and Active Data Guard.

Multi Instance Redo apply is initiated using the optional instances attribute.

SQL> Recover managed standby database disconnect using instances 3;

The optional instances attribute initiates the apply process on all the instances on the standby that are either in mount or Open mode. This same capability can be initiated via the Data Guard broker’s ‘ApplyInstances’ property.

How does this work?

Log shipping works in the same way as previous releases, so the rest of the discussion will be focused on “apply”.  With this new feature, a new Coordinator process will coordinate with the MRP process running on each standby site thereby enabling MRP process from each Standby instance to participate in Redo application.

As shown in the image below, Each standby instance has a MRP process that applies redo on the Standby Site.

It is worth noting that the instances attribute of the "Recover Managed standby database disconnect using instances X" can be different than the actual number of Oracle RAC Standby instances. For example: A setup with 4 Oracle RAC nodes on Primary and 3 Oracle RAC nodes on standby can choose to utilize all 3 standby nodes or just 2 standby instances or even 1 standby instance for redo apply.