Thread: Additional global stats

Here's a patch to add some more data to pg_stat_database. Turns out it
was a lot easier than I thought - especially after reading Alvaros mail.
Luckily I think he was wrong this time ;-)

Does this seem like a reasonable way to do it, or did I miss something
completely?

//Magnus

Magnus Hagander <magnus@hagander.net> writes:
> Here's a patch to add some more data to pg_stat_database. Turns out it
> was a lot easier than I thought - especially after reading Alvaros mail.
> Luckily I think he was wrong this time ;-)

This seems like the wrong way to go about it.  Why not compute these sums
on demand by scanning the stats tables, instead of bloating the stats
collector traffic even more?

            regards, tom lane

> > Here's a patch to add some more data to pg_stat_database. Turns out it
> > was a lot easier than I thought - especially after reading Alvaros mail.
> > Luckily I think he was wrong this time ;-)
>
> This seems like the wrong way to go about it.  Why not compute these sums
> on demand by scanning the stats tables, instead of bloating the stats
> collector traffic even more?
>

Stats traffic should not be affected at all. It just user the same messages already there to increase a couple of
globalcounters per database. The only thing  
that increases is that the stats file gets another 5 counters per database. Given how few databases you normally have
comparedto number of tables, I would  
think that should be pretty trivial.

Scanning them on demand will require loading the whole stats file into the backend instead of just the current
database,which I considered more bloated.. 

/Magnus

"Magnus Hagander" <magnus@hagander.net> writes:
> Scanning them on demand will require loading the whole stats file into
> the backend instead of just the current database,

Good point --- objection withdrawn.

            regards, tom lane

Tom Lane wrote:
> "Magnus Hagander" <magnus@hagander.net> writes:
>> Scanning them on demand will require loading the whole stats file into
>> the backend instead of just the current database,
>
> Good point --- objection withdrawn.

Thanks. I also realized I had not documented the functions, and only the
view. And of course also needed to update the expected files for the
rules regression test. All done ;-)

//Magnus