On Tue, May 7, 2019 at 6:08 AM Robert Haas <robertmhaas@gmail.com> wrote:
> On Mon, May 6, 2019 at 1:58 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > Robert Haas <robertmhaas@gmail.com> writes:
> > > Right. That's why I favor applying the change to move DSM cleanup to
> > > the end for now, and seeing how that goes. It could be that we'll
> > > eventually discover that doing it before all of the AtEOXact_BLAH
> > > functions have had a short at doing their thing is still too early,
> > > but the only concrete problem that we know about right now can be
> > > solved by this much-less-invasive change.
> >
> > But Amit's results say that this *doesn't* fix the problem that we know
> > about. I suspect the reason is exactly that we need to run AtEOXact_Files
> > or the like before closing DSM. But we should get some Windows developer
> > to trace through this and identify the cause for-sure before we go
> > designing an invasive fix.
>
> Huh, OK.
The reason the patch didn't solve the problem is that
AtEOXact_Parallel() calls DestroyParallelContext(). So DSM segments
that happen to belong to ParallelContext objects are already gone by
the time resowner.c gets involved.
--
Thomas Munro
https://enterprisedb.com