Re: Default gucs for EXPLAIN

On Sat, May 23, 2020 at 06:33:48PM +0200, Vik Fearing wrote:
> > Do we really want default_explain_analyze ?
> > It sounds like bad news that EXPLAIN DELETE might or might not remove rows
> > depending on the state of a variable.
>
> I have had sessions where not using ANALYZE was the exception, not the
> rule. I would much prefer to type EXPLAIN (ANALYZE OFF) in those cases.

I suggest that such sessions are themselves exceptional.

> > I think this should be split into two patches:
> > One to make the default explain options configurable, and a separate patch to
> > change the defaults.
>
> This patch does not change the defaults, so I'm not sure what you mean here?

Sorry, ignore that; I wrote it before digesting the patch.

On Sat, May 23, 2020 at 06:16:25PM +0000, Nikolay Samokhvalov wrote:
> Not having BUFFERS enabled by default means that in most cases, when asking
> for help, people send execution plans without buffers info.

I also presumed that's where this patch was going to lead, but it doesn't
actually change the default. So doesn't address that, except that if someone
reports a performance problem, we can tell them to run:

|alter system set explain_buffers=on; SELECT pg_reload_conf()

..which is no better, except that it would also affect any *additional* problem
reports which might be made from that cluster.

If you want to change the default, I think that should be a separate patch/thread.

--
Justin