On 2014-09-27 10:12:03 -0400, Tom Lane wrote:
> > Escaping from blocked send() by pg_terminate_backend().
>
> > I've had a look, but I'd like to have a second opinion on this.
>
> I concur with your opinion that this is scary as heck. We need multiple
> pairs of eyeballs if we're going to do anything in this area. In the long
> run though, I think pushing functionality into signal handlers is exactly
> backwards; we ought to be trying to go the other way.
I very much agree with that concern. The interactions are just
complicated.
I'm now refreshing my work to do all waiting in client communication via
latches. While far from trivial, I'm pretty sure that's the better
course in the long run.
I guess I'll post it in the other thread.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services