Re: Frequent log switches during sqlldr batch jobs

>OK, so in a slightly less bad case, let's say that some calamity
causes
>some disk files to be lost. Is there any reason to think the online
redo
>logs are more likely to be the ones that are lost than the archived
redo
>logs?

Depends on configuration.

>There seems to be an assumption that transactions are "safer" once
>the log file in which are committed is archived. I can see how that
would
>be the case in the long term, when arhived logs are backup up to tape
(but
>in the long term, all logs become archived, so that kind of cancels
except
>at the moment before the copy-to-tape is initiated.). I can see how
that
>would be the case if online logs are duplexed and mirrored while
archived
logs are tetraplexed and mirrored (or something like that). But if arhived
>logs are treated with the same redundancy as online logs, is it true
that
>transactions are safer once the corresponding log has been archived?

Yes. The online redo's are the Achille's heel of Oracle. That is why they should have greater redundancy than anything else. (here is a post that probably makes it more clear:
http://groups-beta.google.com/group/comp.databases.oracle.server/msg/9f6a0d81eed6dad3?dmode=source )

So for one example, if your media crashes and you lose "only" all your archived logs, no problem. If your media with the redo crashes, you've maybe lost whatever transactions were in the active online redo log (or maybe not, that's what ORA-354 is all about, ARCH may be able to keep trying until it can read a proper redo header). If your instance crashes because the media with the redo on it crashed, you will be living in interesting times.

jg

--
@home.com is bogus.
"Sam R. Alapati, OCP (Los Colinas, TX) is an experienced O"  - totality
of Author Biography on Oracle Press online page.