Recovering an instance requires the use of the redo logs, described in Chapter 2. The
redo logs contain a recording of all the physical changes made to the database as a
result of transactional activity, both committed and uncommitted.
The introduction of delayed redo log writes for committed transactions
as an option in Oracle Database 11g could create a scenario
where a transaction appears to have been committed but could still
not be recovered.
The checkpoint concept, also described in Chapter 2, is critical to understanding
crash recovery. When a transaction is committed, Oracle writes all associated database
block changes to the current online redo log. The actual database blocks may
have already been flushed to disk, or may be flushed at some later point. This means
that the online redo log can contain changes not yet reflected in the actual database
blocks stored in the datafiles. Oracle periodically ensures that the data blocks in the
datafiles on disk are synchronized with the redo log to reflect all the committed
changes up to a point in time.
Pages:
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569