Re: shutdown immediate or abort for cold-backup?

On Tue, 28 Aug 2001, Scorpion wrote:

> OK - you are all assuming 1 thing----and you can't say that dropping
> the floor out from under Oracle will not cause corruption. Assuming
> everything goes just right, then the backup and redos will be fine.
> But if just 1 thing happens wrong -----
> Willing to take the chance? Why? This is a MOOT question - hot
> backups, archive logs and a standby database in automatic mode and say
> hello to that magic 99.999% uptime!

The concept of journaling or keeping a transaction log is now used by databases and filesystems from many vendors outside Oracle.

If such technology were not useful for crash recovery, then you would not see the concepts integrated into Veritas VxFS, IBM JFS, SGI XFS, Sun UFS, ReiserFS, ext3, or Oracle, Sybase, DB2, etc. All these systems implement a journal to provide crash recovery, and they are very good at what they do.

While I am no expert in journaling, before any action is begun in manipulating the real data, a plan of the manipulation is recorded into an "intent/transaction/redo log," and only after the intent log is completely written will data modification occur. After a crash, incomplete intent log entries will be erased, and complete intent logs will be checked and "rolled forward" if necessary, bringing the data up to date. Yes, block corruption may occur during a crash, but this "double-entry" technique gives the best assurance that the intact data exists somewhere that SMON can find it.

In the case of Oracle, it seems from the documentation that the only possible case where a shutdown abort could result in an unusable database is if the abort took place during a checkpoint resulting in datafiles with differing checkpoint sequence numbers.

It is HJR's assertion that, assuming that the media is intact, the online redo will contain enough data to restore the system, requiring at most a "recover automatic database" from the administrator (although perhaps SMON takes care of the problem quietly).

Have you any knowledge of this PARTICULAR case, or any case where shutdown abort resulted in an unusable database that did not involve media failure?

   / Charles J. Fisher                   | "Waste no more time arguing what  /
  /  cfisher_at_rhadmin.org                 |  a good man should be. Be one."  /
 /   http://rhadmin.org                  |   -Marcus Aurelius              /
---------------------------------------------------------------------------