At Fri, 13 Dec 2019 13:05:41 +0800, Craig Ringer <craig@2ndquadrant.com> wrote in
> On Wed, 11 Dec 2019 at 02:08, Alvaro Herrera <alvherre@2ndquadrant.com>
> wrote:
>
> > On 2019-Dec-10, Tomas Vondra wrote:
> >
> > > On Tue, Dec 10, 2019 at 09:42:17AM +0900, Kyotaro Horiguchi wrote:
> > > > At Tue, 10 Dec 2019 00:44:09 +0100, Tomas Vondra <
> > tomas.vondra@2ndquadrant.com> wrote in
> >
> > > > I'm not sure how much xact_start for walsender is useful and we really
> > > > is not running a statement there. Also autovac launcher starts
> > > > transaction without a valid statement timestamp perhaps for the same
> > > > reason.
> > >
> > > Maybe, but then maybe we should change it so that we don't report any
> > > timestamps for such processes.
> >
> > Yeah, I think we should to that.
>
>
> Agreed. Don't report a transaction start timestamp at all if we're not in a
> read/write txn in the walsender, which we should never be when using a
> historic snapshot.
>
> It's not interesting or relevant.
>
> Reporting the commit timestamp of the current or last-processed xact would
> likely just be fonfusing. I'd rather see that in pg_stat_replication if
> we're going to show it, that way we can label it usefully.
Sounds reasonable. By the way, the starting of this thread is a valid
value in xact_timestample for a moment at the starting of logical
replication. (I couln't see it unless I inserted a sleep() in
IndentifySystem()). I'm not sure but AFAIS it is the only instance in
walsendeer. Should we take the trouble to stop that? (I put -1 for it)
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center