On Wed, Dec 8, 2021 at 7:34 AM Tomas Vondra
<tomas.vondra@enterprisedb.com> wrote:
> >> I agree it might be useful to provide information about the nature of
> >> the checkpoint, and perhaps even PID of the backend that triggered it
> >> (although that may be tricky, if the backend terminates).
> >
> > Thanks. I agree to have pg_stat_progress_checkpoint and yes PID of the
> > triggered backend can possibly go there (we can mention in the
> > documentation that the backend that triggered the checkpoint can get
> > terminated).
> >
>
> My concern is someone might run something that requires a checkpoint, so
> we start it and put the PID into the catalog. And then the person aborts
> the command and starts doing something else. But that does not abort the
> checkpoint, but the backend now runs something that doesn't requite
> checkpoint, which is rather confusing.
>
> >> I'm not sure about adding it to control data, though. That doesn't seem
> >> like a very good match for something that's mostly for monitoring.
> >
> > Having it in the control data file (along with the existing checkpoint
> > information) will be helpful to know what was the last checkpoint
> > information and we can use the existing pg_control_checkpoint function
> > or the tool to emit that info. I plan to add an int16 flag as
> > suggested by Justin in this thread and come up with a patch.
> >
>
> OK, although I'm not sure it's all that useful (if we have that in some
> sort of system view).
If the server is down, the control file will help. Since we already
have the other checkpoint info in the control file, it's much more
useful and sensible to have this extra piece of missing information
(checkpoint kind) there.
> >> We already have some checkpoint info in pg_stat_bgwriter, but that's
> >> just aggregated data, not very convenient for info about the current
> >> checkpoint. So maybe we should add pg_stat_progress_checkpoint, showing
> >> various info about the current checkpoint?
> >
> > +1 to have pg_stat_progress_checkpoint view to know what's going on
> > with the current checkpoint.
> >
>
> Do you plan to add it to this patch, or should it be a separate patch?
No, I will put some more thoughts around it and start a new thread.
Regards,
Bharath Rupireddy.