Thanks!
> > 2/
> > I don't see we have tests for other timeout based GUCs, but it would nice
> > to ensure that this woks correctly. Maybe as a custom_stats test where we
> > SET stats_flush_interval inside the transaction and make sure the stats flush
> > only after the new timeout has passed. Maybe?
>
> Not sure I follow, that's what 0002 is doing.
I was referring to "SET stats_flush_interval", which you are doing in 0005,
so disregard this comment. I wrote this comment when I read 0003 only.
> > Will this be tue at all time? Let's imagine a Kind that flushes all the fields
> > ANYTIME, would we not want to delete the pending entry?
> >
> > + /* if successfull non-partial flush, remove entry */
> > + if (did_flush && !is_partial_flush)
> > pgstat_delete_pending_entry(entry_ref);
> >
>
> Right, and that's why the MIXED flush mode was useful, i.e to be able to distinguish
> here. So, in the attached, instead of re-introducing the MIXED flush mode, I added
> a "is_partial" bool paramater to the flush_pending_cb().
>
Yeah, MIXED flush mode to deal with this is still not the right idea, IMO.
I do agree with what you did here by passing "is_partial" to the callbacks.
Now it will be up to the callbacks to determine if they performed a partial
or a complete flush ( even in the case of any anytime flush, if they so
happened to flush stats for all the fields ).
I do not have any further comments on this patchset.
--
Sami Imseih
Amazon Web Services (AWS)
