Recovery Manager Restore and Recovery Techniques - Oracle 10g/11g Data and Database Management Utilities - page 124

The controlfile is a critical file during the recovery process as this file contains the
information that marks the transactional point the database must reach to perform a
complete recovery process.
Instance failure
An instance failure happens when the database did not receive the last checkpoint
before the instance was shutdown, leaving the datafiles in an inconsistent state. This
can happen when a SHUTDOWN ABORT command is issued or when the host is abruptly
shutdown after a power failure and there was no way to properly shutdown the
database with a SHUTDOWN NORMAL|IMMEDIATE|TRANSACTIONAL command. At
startup time Oracle realizes there is a difference between the SCN information stored
in the controlfile and the SCN information recorded at the datafile headers. Once
the difference has been stated the System Monitor Background Process ( SMON )
starts an automatic instance recovery process by reading the missing transactions
from the redo log files and applying them against the datafiles requiring them. This
process does not require DBA intervention and it does not require the database to
be in archivelog mode. Once all the transactions have been applied the database is
automatically opened by the instance.
Media failure
A media failure happens when a datafile gets corrupted or lost. This requires the
DBA to take a valid backup and restore the missing datafile or datafiles to their
original location (if possible) or to a different location. The last applied transaction
found in the restored datafiles comes from the backup time, which most probably
is far away from the range stored at the redo log files, so it is required to read the
missing transactions from the archivelog files. At this point the startup process will
be stopped thus letting the DBA know that a database recovery process is required.
If the database is not in Archivelog mode then it is not possible
to apply a database recovery process and in the best case
scenario the only information that will be restored will be from
the last valid offline backup time.
In the following image a recovery process takes place:
1. The first step restores the missing datafiles from a valid backup.
2. At this point all physical structures are available but there are missing
changes, so a recovery process is required.
3. Once the process is finished the database is open.