Re: Resetting spilled txn statistics in pg_stat_replication

On Tue, 2 Jun 2020 at 17:22, Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> wrote:
>
>
>
> On 2020/06/02 16:00, Kyotaro Horiguchi wrote:
> > At Tue, 2 Jun 2020 15:17:36 +0900, Masahiko Sawada <masahiko(dot)sawada(at)2ndquadrant(dot)com> wrote in
> >> Hi all,
> >>
> >> Tracking of spilled transactions has been introduced to PG13. These
> >> new statistics values, spill_txns, spill_count, and spill_bytes, are
> >> cumulative total values unlike other statistics values in
> >> pg_stat_replication.
>
> Basically I don't think it's good design to mix dynamic and collected
> stats in one view. It might be better to separate them into different
> new stats view. It's too late to add new stats view for v13, though....

Yeah, actually I had the same impression when studying this feature.
Another benefit of having a separate view for such statistics would be
that it can also support logical decoding invoked by SQL interface.
Currently, reorder buffer always tracks statistics of spilled
transactions but we can see it only in using logical replication.

Regards,

--
Masahiko Sawada http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services