I am new to RAC so I am trying to come to terms with the scope of what it takes to support it. I am using Oracle 9i on Sun Solaris 8 with Sun's clusterware. Could you please list the areas that a DBA must address both to create and support a RAC environment that would not otherwise exist in a traditional Oracle database?
I know some would include:
Hardware Realm
* A computer system is required for each node.
* Each system must have Fiber cards connecting to the clustered file system solution.
* Cluster software installation.
* Cluster software maintenance.
DBA Realm
* The creation of new tablespaces requires a corresponding cluster\raw file system piece created.
* The creation of new control and redo files requires a corresponding cluster\raw file system piece created.
* In a DSS environment that requires very proactive partition maintenance the creation of partition data files requires a corresponding cluster\raw file system piece created.
* Archive logs cannot reside on a clustered file system.
* Though all instances can write to one disk, every instance generates it's own archive logs.
* RMAN backups files cannot reside on a clustered file system (it would not be practical).
* There are special considerations concerning backups, restoration and recovery.
* If unchecked, control or other files can grow larger than the cluster\raw allotted space which can cause database failure in some instances (this has happened for me).
* Applying common Oracle patches requires special considerations.
Other Considerations
* If the database is down so are all nodes. Example: If the SYSTEM tablespace requires restoration RAC does not provide database fail over (only instance).
What am I missing?
Thanks,
Michael42
I know some would include:
Hardware Realm
* A computer system is required for each node.
* Each system must have Fiber cards connecting to the clustered file system solution.
* Cluster software installation.
* Cluster software maintenance.
DBA Realm
* The creation of new tablespaces requires a corresponding cluster\raw file system piece created.
* The creation of new control and redo files requires a corresponding cluster\raw file system piece created.
* In a DSS environment that requires very proactive partition maintenance the creation of partition data files requires a corresponding cluster\raw file system piece created.
* Archive logs cannot reside on a clustered file system.
* Though all instances can write to one disk, every instance generates it's own archive logs.
* RMAN backups files cannot reside on a clustered file system (it would not be practical).
* There are special considerations concerning backups, restoration and recovery.
* If unchecked, control or other files can grow larger than the cluster\raw allotted space which can cause database failure in some instances (this has happened for me).
* Applying common Oracle patches requires special considerations.
Other Considerations
* If the database is down so are all nodes. Example: If the SYSTEM tablespace requires restoration RAC does not provide database fail over (only instance).
What am I missing?
Thanks,
Michael42