Oracle RAC with single node Data Guard physical standby

Question

Best practices for Oracle 11gR2 Data Guard implies that the primary and standby databases should have the same layout, eg. RAC to RAC.

What special considerations would one need to take into account to replicate a RAC database to a single node physical standby?

Any reason why this would be a very bad idea?

Answer 1

There aren't any special considerations that need to be taken into account when setting up a RAC->No RAC Primary->Standby configuration.

In fact, there's a White Paper written by Oracle which explains the setup.

The Oracle documentation is here.

Answer 2

In my environment I only found a few additional init parameters that needed changing, compared to a single instance to single instance configuration.

These had to be removed or altered when creating the initial pfile for the standby instance.

This one is rather obvious, and must be set to false:

cluster_database = false

These should probably not be configured on the single instance standby:

remote_listener
listener_networks

Similarly, if we're using RMAN to create the standby database, the clone script need to address the same parameters and could look something like this:

DUPLICATE TARGET DATABASE
  FOR STANDBY
  FROM ACTIVE DATABASE
  DORECOVER
  SPFILE
    SET DB_UNIQUE_NAME='STBY' COMMENT 'standby'
    SET LOG_ARCHIVE_DEST_2='SERVICE=PROD ASYNC VALID_FOR=(ONLINE_LOGFILES,PRIMARY_ROLE) DB_UNIQUE_NAME=PROD'
    SET FAL_SERVER='PROD' COMMENT 'primary'
    SET CLUSTER_DATABASE='FALSE' COMMENT 'standby is single instance'
    SET REMOTE_LISTENER='' COMMENT 'standby is single instance'
    SET LISTENER_NETWORKS='' COMMENT 'standby is single instance'
 NOFILENAMECHECK;

There is a reasonable overview in the 10g MAA white paper mentioned by Phil. I couldn't find much in the 11g MAA white paper, but it can be worked out from the data guard documentation or simply from trial and error.

Also, I was unable to find references to using the SCAN listener of the clustered primary, so I let the standby instance connect through a dedicated VIP listener, which should be fine. I suppose one might not want data guard traffic going over the public network anyway.