Re: Delay locking partitions during INSERT and UPDATE

On Fri, Feb 1, 2019 at 9:16 AM David Rowley
<david(dot)rowley(at)2ndquadrant(dot)com> wrote:
> On Sat, 2 Feb 2019 at 03:07, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> > I'm now wondering whether the same issues discussed in
> > https://www.postgresql.org/message-id/CA%2BTgmoZN-80143F8OhN8Cn5-uDae5miLYVwMapAuc%2B7%2BZ7pyNg%40mail.gmail.com
> > also need discussion with respect to this patch. But I haven't
> > thought about it very hard, so I'm not sure whether they do or don't.
>
> I really don't think it does, or if it does then the code is already
> broken as it is now.
>
> As the code is today, we obtain the locks well after the plan is
> locked in. The only difference with this patch is that we do the
> locking on routing tuples to the partition for the first time rather
> than do them all at once when setting up the routing data structure.

OK, that is what I thought yesterday, and then I was just doubting
myself. Thanks for thinking about it.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company