Re: standbys and unrecoverable operations

unrecoverable_change is from v$datafile, which is from controlfile, not from datafile header.

if you refresh controlfile at standby, this all the information at standby v$datafile is incorrect.

It is very hard to detect. in a complex production environment.

enable force_logging in 9i is a good idea for standby

On 11/16/05, Bobak, Mark <Mark.Bobak_at_il.proquest.com> wrote:
> Josh,
>
> I'm not a standby or DataGuard expert, but, think about what the
> unrecoverable change # represents. It's the SCN at which the last
> unrecoverable operation occurred on that datafile. So, if the primary is
> ahead of the standby, that means there are operations which have occurred on
> the primary which did not propogate to the standby. This is a corruption
> waiting to happen. If you activate the standby and a datablock is accessed
> that was loaded unrecoverable on the primary, you'll encounter an ORA-26040.
>
> So, I assume (don't have a standby setup handy to confirm it) that the
> default safe position is that they are equal, since, when you clone from the
> primary to initially create the standby, they'd (presumably) be equal. At
> that point, the only way for them to get out of sync is if you do an
> unrecoverable (aka nologging) load in the primary database.
>
> Corrections welcome from those with actual DG and standby experience! ;-)
>
> Hope that helps,
>
> -Mark
>
> PS Note that in 9i (can't remember if it was 9.0.1 or 9.2.0 intorduction)
> you can do ALTER DATABASE FORCE_LOGGING=TRUE; and everything will log, even
> if people try to do nologging loads.
> ________________________________
> From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
> On Behalf Of Josh Collier
> Sent: Tuesday, November 15, 2005 4:52 PM
> To: oracle-l_at_freelists.org
> Subject: standbys and unrecoverable operations
>
>
> Greetings,
>
> The Oracle documentation says that if the unrecoverable_change# for a
> datafile reported (v$datafile) by the primary is greater than that reported
> by the standby then you will need to recover that datafile (by copying it
> over from the primary) in order to avoid block corruption errors if the
> standby is activated.
>
> Does this also hold if the unrecoverable_change# are identical?
>
> have a good day,
>
> Josh C.

--
Regards
Zhu Chao
www.cnoug.org
--
http://www.freelists.org/webpage/oracle-l