At Fri, 28 Jan 2022 10:00:46 +0800, Julien Rouhaud <rjuju123@gmail.com> wrote in
> Hi,
>
> On Fri, Jan 28, 2022 at 10:38:53AM +0900, Kyotaro Horiguchi wrote:
> >
> > I'd like to see the PID of the triggering process, but it is really
> > not a information suitable in the control file...
>
> Yes that's something I would like too. But even if the PIDs could be store, I
> don't think that having the information for an already completed checkpoint
> would be of any use at all.
>
> For the current checkpoint, it should also be an array of PID. For instance if
> the checkpointer started a throttled checkpoint, then someone calls a non
> immediate pg_start_backup() and finally thinks it's too slow and need a fast
> checkpoint. This would be welcome in a new pg_stat_progress_checkpoint view.
Yeah, I thought of the same issue. But at the time of the previous
mail I thought it would be enough that the PID is of the first
initiator of the current running checkpoint likewise the checkpoint
kind. So it is important to think about the use case. If we put
significancy on actually happend checkpoints, we don't need absorbed
checkpoint requests to be recorded but if we put significance on
checkpoint requests, we need to care of every checkpoint request.
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center