> The duration suddenly goes from 270 milliseconds to 173 seconds! The index
> scan on bprofile_comments_status_idx suddenly shows 15288 loops, where it
> should be 1 loop just like before. So shomehow the 9.0 planner gets it all
> wrong.
>
> I also noticed that normally I get an iowait with a
* Uwe Schroeder (u...@oss4u.com) wrote:
> Yes, the database is vacuumed and analyzed. The bad plan from 9.0 improves by
> 2 seconds when I go for a really high statistics target of 5000.
What if you go back to 10..?
Stephen
signature.asc
Description: Digital signature
> * Uwe Schroeder (u...@oss4u.com) wrote:
> > Now I turn off the 8.3 instance and start the 9.0 instance. Remember,
everything is identical. Here the same query again:
> Everything isn't identical if you just started PG 9.0 though- presumably
> the 8.3 instance had everything cache'd already. W
> On 29 January 2011 09:11, Uwe Schroeder wrote:
> > Maybe someone here can make sense of this.
> > I'm trying to upgrade a 8.3 system to a 9.0 system. Usual procedure dump,
> > restore, vac full, reindex.
> >
> > Both - old and new - run on the same hardware and the postgresql.conf
> > setting
On 29 January 2011 09:11, Uwe Schroeder wrote:
>
> Maybe someone here can make sense of this.
> I'm trying to upgrade a 8.3 system to a 9.0 system. Usual procedure dump,
> restore, vac full, reindex.
>
> Both - old and new - run on the same hardware and the postgresql.conf
> settings are identic
* Uwe Schroeder (u...@oss4u.com) wrote:
> Now I turn off the 8.3 instance and start the 9.0 instance. Remember,
> everything is identical. Here the same query again:
Everything isn't identical if you just started PG 9.0 though- presumably
the 8.3 instance had everything cache'd already. What hap
