Graceful Switchover and RAC

Graceful Switchover and RAC

Graceful Switchover can be used with Oracle RAC and works identical to graceful switchover with single instance databases. However the following applies:

1. Graceful Switchover can only occur between ONE primary node and the standby server. The instance on the other primary node(s) will be automatically shutdown during graceful switchover.

For example: With the following configurations:

Primary

Standby

Primary

Standby

ServerA – RAC instance

ServerC – Single instance standby database

ServerB – RAC instance

 

Graceful Switchover with RAC example

With the following configuration:

Primary

Standby

Primary

Standby

ServerA – RAC instance dbvrac1

ServerC – Single instance standby database dbvrac

ServerB – RAC instance dbvrac2

 

Graceful Switchover is initiated on both ServerA and ServerC with command:


The following will occur:

1. RAC database dbvrac2 on ServerB will be shutdown.

2. RAC database dbvrac1 will be converted to a single instance standby database.

 

3. Single instance standby database on ServerC will be converted to a single instance primary database with 2 threads:

Standby

Primary

Standby

Primary

ServerA – Single instance standby database dbvrac1

ServerC – Single instance primary database dbvrac

ServerB – RAC instance shutdown

If Graceful Switchover is initiated again to bring it back to the original configuration the following will occur:

 

1. Single instance standby database dbvrac1 on ServerA will be converted to a single instance primary database with 2 threads.

 

2. Single instance primary database dbvracon ServerC will be converted to a single instance standby database with 2 threads:

Primary

Standby

Primary

Standby

ServerA – Single instance primary database dbvrac1

ServerC – Single instance standby database dbvrac

ServerB – RAC instance shutdown

 

To convert from a single instance primary database back to a RAC database:

 

1. On ServerA, set database parameter cluster_database=TRUE in database dbvrac1.

2. Restart instance dbvrac1 on ServerA.

3. Start dbvrac2 instance on ServerB.

4. Dbvisit Standby can be initiated again as normal on ServerA and ServerC:


5. On other nodes that were not part of the graceful switchover (in this case ServerB), Dbvisit Standby must be resynchronised with the standby database by running command:

The command only needs to run once. After this command, Dbvisit Standby can be run as normal: