Many new features have been introduced into the Recovery Manager, of which but a few are:
Persistent
configurations that can be created and applied to any session.
This includes automatic channel configuration.
Retention policy features that you can use to create a moving
"recovery windows" for your database.
Automatic control file backup and restore means that
you can restore a control file even of the current control
file and
RMAN catalogue are lost.
Block Media recovery to perform recovery of particular blocks
in a datafile whilst that datafile is still online. Comes with
a new dynamic view v$_database_block_corruption.
Enhanced archive log backup features include the ability to
switch log files before backing up the logs to produce a consistent
state as you would in a self-scripted backup procedure.
Log
Miner Enhancements:
The Log Miner from 8i has been significantly enhanced to include the ability
to view DDL statements, direct path inserts and data dictionary information.
In addition extra processing features include the ability to view collections
of DML statements as transactions; place complex 'where clauses' on the information
you are extracting so you can see particular classes of statements and generate SQL
redo and undo with primary key information to help the DBA rollback changes.
All this is finished off with a brand new GUI, the Log Miner Viewer.
Flashback Queries:
By setting a new init.ora parameter
(and having enough disk space!) you can cause the database to retain redo/undo information
for a certain period of time. Then with the help of a new PL/SQL package, DBMS_FLASHBACK,
you can retrieve old images of data in the specified timeframe and restore if necessary.
Requires that you use the new automatic undo management feature.
"alter database end backup" command:
Takes
all datafiles out of backup mode in a single command. This
new command allows you to script database crash
recovery procedures without the need for user intervention.
Trial recoveries:
You can now practice applying redo to the database in memory without ever writing the information to disk by using the SQL*Plus "recover…test" command. This means that you can foresee potential recovery problems.
Fast Fault Recovery:
A new parameter,
FAST_START_MTTR_TARGET, allows you to specify an expected mean time to recover. Oracle
then automatically uses this value to set dirty buffer limits and limit the amount of
redo generated since the last checkpoint by adjusting the rate at which information
is written.