Re: OPS Backup/Restore/Recovery

Rich Woods wrote:
>
>
> For backup purposes, it does matter that this is an OPS environment as
> you've two instances that have separate control files and redo threads.
> You need both threads of redo in order to perform recovery.
> --
> Rich Woods
> Technical Field Support Specialist, Oracle Corporation, USA
> The above statements and opinions are my own and do not
> necessarily represent those of Oracle Corporation.

In an Oracle Parallel Server environment, it does not matter, each instance has
equal access to all of the shared raw devices that make up the database. It is
true that both threads of the redo are required to perform recovery, however,
one or all of the instances of the parallel server must have access to the redo
logs of all of the threads otherwise recovery will not be possible. Recovery of a
OPS database is handled through one instance only.

There also can only be one virtual control file in an Oracle database. Even in an OPS
environment. Multiple control files are recommended to prevent the loss of the
database in the event that a disk housing the control file is lost. In an OPS
environment, these control files should/can be located on each of the nodes making
up the OPS environment.

By the way, getting back to the original question, Sql*Backtrack is a great product
for backing up such an environment. We have an 8-node MPP OPS environment > 200GB
and we have not had any problems with the tool. All of the backup/recovery activity
is handled through a single node (instance).

Patrick Chambers
Principal Consultant
DB Systems Integration
dbsysint_at_interlynx.net Received on Tue Mar 11 1997 - 00:00:00 CST