instance recovery and media recovery

The essential difference between crash/instance recovery and media recovery is:

Crash/instance recovery applies the redo log from the incremental checkpoint (incremental checkpoint) for the instance that needs to be recovered. Because the first condition of log coverage is to complete the logfile switch checkpoint of the relevant log, and the archive log is always behind the checkpoint of the instance by definition, so for crash/instance Recovery crashes or instance restores always require only access to read-online redo logs (on-line redo logfile).

Media recovery from the checkpoint of the old data file apply Redo reference redo logs, which may have originated from backup. The archive redo log is required for media recovery, so Rman or DBA (user-managed backup) also needs to restore the archived logs associated with the backup.

Crash/instance recovery always guarantees that recovery is complete only after all persistent redo data has been applied. It is considered that the crash instance completes the recovery process when Oracle guarantees that all committed transactions have been included for recovery.

Instead, the media recovery has a reference to incomplete recovery (incomplete recovery) and partial recovery (partial recovery) to achieve consistency of the recovery database (db) to a point in time.

The similarities between crash/instance recovery and media recovery are:

crash/instance recovery and media recovery are both traditional roll-forward recovery methods (rolling forward), which are in principle a repetition of persistent redo log data. Whether crash/instance recovery or media recovery roll forward, a transaction rollback is required to roll back uncommitted transactions, although the database can be opened without waiting for rollback to complete after the roll forward. But we think the database is really consistent only when the rollback is complete.

Personal Understanding:

In literal terms, instance recovery is instance recovery The reason is instance failure, media recovery medium reason is media failure

Instance recovery means that the database fails (instance crash), or shutdown Abort,startup Force, the buffer Chache data is not written to the data file, which means

Data file's head SCN number and the current instance of the SCN number is inconsistent, need to roll forward from the current redo, to ensure that buffer cache data written to the data file, and the database in the strength before the collapse of the state, instance recovery is generally a database restart when the automatic application redo, Typically done by Smon, no user intervention required

Media recovery means that the storage media is corrupted and must be shown to restore the database from the backup or archive log, such as recover Database,recover tablespace, recover datafile

It is possible to perform incomplete or partial recovery, not necessarily to restore the latest SCN number to the database