Re: Another crazy hot backup theory in a book

On Thu, 24 Sep 1998, Keith Boulton wrote:
>
> You are being unreasonably picky. For most users except fulltime
> Oracle DBAs, the description given in both your examples is adequate
> and logically correct.

You may be right that I am being too picky for the purposes of a SQL developer, but anyone who is planning to do hot backups needs to understand that datafiles are, in fact, written during hot backup. The statement that writes are "diverted" to the redologs is not logically correct by any means. They should have said that full block images of datafile writes are logged during backup mode. To suggest that the writes are somehow "diverted" is just wrong.

This is important to understand for even a shadetree DBA. Take the example of a hard instance crash during hot backup. If a DBA thinks that datafiles are not written during hot backup, she will think that she must apply archived redo for the entire duration of the tablespace backup in order not to lose data. In fact, she could just say "alter database datafile... endbackup;" and no data will be lost. Also, the instance will be able to proceed immediately to crash recovery of the most recent online redolog.

I don't think it is unreasonably picky to expect that these prolific Oracle books be accurate. If the theory behind hot backups is outside of their scope, they shouldn't flail around trying to explain it, then get it wrong anyway.

YMMV :-) --
Jeremiah Wilton http://www.wolfenet.com/~jeremiah Received on Thu Sep 24 1998 - 13:31:23 CDT