Date: Wed, 9 Sep 2020 14:06:54 -0700
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-hackers@lists.postgresql.org
Subject: Re: SIGQUIT handling, redux
Message-ID: <20200909210654.ayufghpoczz7jfmx@alap3.anarazel.de>
References: <1850884.1599601164@sss.pgh.pa.us>
 <20200909192620.dylomoxknmt6rdb3@alap3.anarazel.de>
 <110116.1599682140@sss.pgh.pa.us>
 <20200909202201.unpjmshshu7sge6i@alap3.anarazel.de>
 <112673.1599683437@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <112673.1599683437@sss.pgh.pa.us>
Precedence: bulk

Hi,

On 2020-09-09 16:30:37 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > On 2020-09-09 16:09:00 -0400, Tom Lane wrote:
> >> We could call it startup_packet_die or something?
> 
> > Yea, I think that'd be good.
> 
> I'll make it so.

Thanks!


> >> We see backends going through this code on a very regular basis in the
> >> buildfarm, but complete hangs are rare as can be.  I think you
> >> overestimate the severity of the problem.
> 
> > I don't think the BF exercises the problmetic paths to a significant
> > degree. It's mostly local socket connections, and where not it's
> > localhost. There's no slow DNS, no more complicated authentication
> > methods, no packet loss. How often do we ever actually end up even
> > getting close to any of the paths but immediate shutdowns?
> 
> Since we're talking about quickdie(), immediate shutdown/crash restart
> is exactly the case of concern, and the buildfarm exercises it all the
> time.

Yea, but only in simple cases. Largely no SSL / kerberos. Largely
untranslated. Mostly the immediate shutdowns aren't when inside plpython
or such.


> > And in the
> > SIGQUIT path, how often do we end up in the SIGKILL path, masking
> > potential deadlocks?
> 
> True, we can't really tell that.  I wonder if we should make the
> postmaster emit a log message when it times out and goes to SIGKILL.
> After a few months we could scrape the buildfarm logs and get a
> pretty good handle on it.

I think that'd be a good idea.

Greetings,

Andres Freund