Re: Insert /* +append */ always ?

"Saikat Chakraborty" <saikatchak_at_hotmail.com> wrote in message news:ce53c312828ba02c6f9286612fdfbb6e.16981_at_mygate.mailgate.org...
> "Thomas Kyte" <tkyte_at_oracle.com> wrote in message
> news:addtbq02m2j_at_drn.newsguy.com
>
> >
> > What that means is that the data need no be protected by REDO in
noarchivelog
> > mode. We don't need to redo the operation since the operation is done
directly
> > to disk. Since media recovery is not possible in noarch mode -- no redo
is
> > needed (we don't need the redo for instance recover in these cases since
the
> > blocks are already safely on disk)
> >
> > In archivelog mode, we do need the redo and the operations would have to
be
> > performed on NOLOGGING tables in order to bypass redo generation....
> >
>
> Hi Tom,
>
> I tested that rows inserted by append hint does appear after
> a database recovery. But I have a question:
> If we have a hot standby database,
> does the standby database also gets those records? If not the standby
> will silently gets corrupted.
> I am talking of Oracle 8i and above versions.
> Thanks for guiding us,
> Saikat

Sorry for jumping in.

The primary database for a Standby database must be running in archivelog mode. So the condition that Howard got above (CTAS is done as an unrecoverable operation) doesn't apply.

If you have actually chosen to perform NOLOGGING operations then you effectively have two choices.

1. Recreate the standby database(s) from scratch. Fairly straightforward but a large waste of time.
2. update the affected datafiles (and the standby control file) from a fresh primary database backup and restart managed recovery. Potentially quicker but you would need to be absolutely sure you got it right.

Obviously you then have choice 3 rethink the rationale for performing NOLOGGING operations on a database where you rely on the redo log stream to maintain a standby.

--
Niall Litchfield
Oracle DBA
Audit Commission UK