Re: Are these too harsh ?

"Norman Dunbar" <ndunbar_at_lynxfinancialsystems.co.uk> wrote in message news:F43E6BAE5BB5D411A44C00805FBE740DAEED0D_at_apps.leeds.lfs.co.uk... [Snip]
>
> PS. Am I right in the following :
>
> Q. How do I allow end-user queries against a standby database without
> damaging its ability to act as a standby?
> A. I don't think you do. Once the standby database is opened, it is no
> longer a standby. So as for reading only from one ...
> Of course, if you mean a read only database (all tablespace in read only
> mode) then that's a different matter. Or have I missed something between
> O7/O8 and the present day as far as stand by databases go ?
>

In 8i, they introduced the concept of opening the standby database is read only mode. Having done your reads, you can resume recovering the standby. That is entirely different from *activating* the database, which turns it into the fully-open, full-blown new production database, and from which there is indeed no return.

New feature in 8i.

Incidentally, as far as Connnor is concerned, I think a more interesting question would be 'I've opened my standby database for read only access, and 98% of my reports are coming out fine. 2% keep failing though. What might the cause be?'

(To which the answer would be: you created your temporary tablespaces using the old Oracle 8 'datafile' terminology, not the new 8i 'tempfile' one... datafiles can't ever be written to in read only mode, tempfiles can. Nice understanding of quite a lot of concepts).

> Q. You issue 'select count(*) from my_table;' and it takes a
> tremendously long time, but then return no rows. What could be causing
> this ?
> A. A shed load of blocks being put back onto the freelists after a
> recent and very large DELETE statement perhaps ?
>

Sort of, I'd say. I guess the terminology Connor might be looking for is... it was an absolutely HUGE table, with a gazillion rows, but I did a massive DELETE from HUGE, which piece of DDL unfortunately leaves the high watermark stranded somewhere at the end of the datafile -and count(*) needs to do a full tablescan, and doesn't finish until it hits the high watermark.

Something like that anyway!

Have fun,
HJR
> ------------------------------------------------------------------------
> --------
> Norman Dunbar EMail: NDunbar_at_LynxFinancialSystems.co.uk
> Database/Unix administrator Phone: 0113 289 6265
> Lynx Financial Systems Ltd. Fax: 0113 201 7265
> URL: http://www.LynxFinancialSystems.com
> ------------------------------------------------------------------------
> --------
>
Received on Wed Jun 27 2001 - 05:13:12 CDT