all the databases of the cluster are regularly vacuumed (at least once a day),
all the stats are up to date.
what i'm talking about is a partitionned table (8.1, with constraint
exclusions), a 'main' table empty, and a dozen of inherited tables, which
contain the real data. The main table contains 0 record, only the inheriting
tables contain data. So the stats are right saying that the main table
contains 0 record. Then, seeing that, pgadmin decides to count the record (as
there aren't supposed to be a lot of them), but gets to read millions of
record (the content of all the partitions).
On Monday 10 April 2006 09:54, Andreas Pflug wrote:
> Marc Cousin wrote:
> > Hi,
> >
> > I've got a partitionned table with PG 8.1, and I'm using pgadmin3 1.4.1.
> > I'm having problems using pgadmin because it tries to count the elements
> > of the partitionned table...
> >
> > The main table contains 0 record (so the stats say 0), so pgadmin decides
> > it can count all the elements from the table... but the count() on this
> > table is avout 40 000 000 record including all the partitions. The
> > consequence is that everytime I click on the main table, pgadmin scans
> > all the partitions (taking about 3 minutes and slowing down the
> > server...)
> >
> > Is this a known problem ?
>
> Yes. You clearly ignored the Guru's hint to vacuum.
>
> Regards,
> Andreas