My question relates to Oracle Data Guard. Why shouldn't we run the Media Recovery Process (MRP) on multi-node Real Application Clusters (RAC) Data Guard. Why only run MRP on one node?
As you know, the Media Recovery Process (MRP) for Data Guard is the process that applies the archived log files to the physical standby database. While the log files may originate from multiple nodes, to ensure consistency, the application of the logs in the recovery process is still a sequential operation, meaning there would be no value to having multiple recovery processes.
If you are looking for ways to speed up the recovery process, there are several options available. Some people have adjusted their parallel_execution_message_size initialization parameter. In many cases adjusting this parameter above the default size has shown great improvement in redo apply performance, at the cost of higher memory utilization.
More info on this parameter is here: http://docs.oracle.com/cd/B28359_01/server.111/b28320/initparams170.htm
Also, more general information on improving your recovery performance can be found in Oracle's MAA Best Practices Documentation, found on page 28 here: http://www.oracle.com/technetwork/database/features/availability/maa-wp-11gr1-activedataguard-1-128199.pdf
This was first published in January 2012