Re: Delay locking partitions during INSERT and UPDATE

On Thu, Jan 24, 2019 at 4:43 PM Tomas Vondra
<tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
> Yes, I don't see why would the patch change that and I've been looking
> for such cases. I think David was looking at that this week too, and I
> assume he'll send an update if he finds anything. If not, I plan to get
> it committed soon-ish (possibly next week after the FOSDEM dev meeting,
> unless there are some objections).

I have reviewed this patch and I am in favor of it. I think it likely
needs minor rebasing because of the heap_open -> table_open renaming.
I also agree that it's worth taking some deadlock risk for the rather
massive performance gain, although I suspect it's likely that a few
users are going to complain about deadlocks and I wonder whether we'll
have to put some energy into that problem at some point. However, I
think what we probably want to do there is reduce the probability of
deadlocks through some trickery or maybe invent some new locking
mechanisms that work around the problem. The alternative of trying to
block this patch seems unpalatable.

In short, +1 from me.

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