Re: Resetting spilled txn statistics in pg_stat_replication

On Tue, Sep 8, 2020 at 7:53 AM Masahiko Sawada
<masahiko(dot)sawada(at)2ndquadrant(dot)com> wrote:
>
> On Mon, 7 Sep 2020 at 15:24, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> I'm still going to work on this patch although I might be slow
> response this month.
>

This is a quite fast response. Thanks for staying on top of it.

>
> >
> > 9. While reviewing this patch, I noticed that in
> > pgstat_read_db_statsfile_timestamp(), if we fail to read ArchiverStats
> > or SLRUStats, we return 'false' from this function but OTOH, if we
> > fail to read 'D' or 'R' message, we will return 'true'. I feel the
> > handling of 'D' and 'R' message is fine because once we read
> > GlobalStats, we can return the stats_timestamp. So the other two
> > stands corrected. I understand that this is not directly related to
> > this patch but if you agree we can do this as a separate patch.
>
> It seems to make sense to me. We can set *ts and then read both
> ArchiverStats and SLRUStats so we can return a valid timestamp even if
> we fail to read.
>

I have started a separate thread for this bug-fix [1] and will
continue reviewing this patch.

[1] - https://www.postgresql.org/message-id/CAA4eK1J3oTJKyVq6v7K4d3jD%2BvtnruG9fHRib6UuWWsrwAR6Aw%40mail.gmail.com

--
With Regards,
Amit Kapila.