On Fri, 2025-04-25 at 12:36 +1200, David Rowley wrote:
> On Fri, 25 Apr 2025 at 03:06, Laurenz Albe <laurenz.albe@cybertec.at> wrote:
> > On Fri, 2025-04-25 at 01:41 +1200, David Rowley wrote:
> > > The 79.3 seconds is the total time spent doing reads for all parallel
> > > workers. 52.6 seconds is the wall clock time elapsed to execute the
> > > query.
> >
> > But wouldn't it read "loops=3" or similar then?
>
> Only if the Nested Loop was below the Gather / Gather Merge node.
Ah, I see. The I/O time accrued in a different, lower, parallelized step
of the execution plan.
Yours,
Laurenz Albe
