From: Jerry Sievers <gsievers19@comcast.net>
To: "Brandl\, Wolfgang" <wolfgang.brandl@brz.gv.at>
Cc: "pgsql-admin\@postgresql.org" <pgsql-admin@postgresql.org>
Subject: Re: restart_after_crash
References: 
 <F2EB158D72DC6A4F97877DD0BB6F020E65065972@CMD0MBX-AB01.cmd0.cna.at>
Date: Wed, 12 Sep 2018 09:43:44 -0500
In-Reply-To: 
 <F2EB158D72DC6A4F97877DD0BB6F020E65065972@CMD0MBX-AB01.cmd0.cna.at>
	(Wolfgang Brandl's message of "Wed, 12 Sep 2018 07:52:40 +0000")
Message-ID: <87lg86vkrj.fsf@jsievers.enova.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

"Brandl, Wolfgang" <wolfgang.brandl@brz.gv.at> writes:

> Hi
> We are testing
> restart_after_crash = on
>
> What is a cras? We killed the Postgres process with -9 SIGSEGV SIGABRT etc.. But never postgres is restarting.
>
> The question arose because of the decision to restart Postgres over systemctl or with the Postgres feature.
>
> But it seems that the Postgres restart is not working or not always working. Or is an additional feature necessary?

It is for a postmaster controlled kill of all backends after a *backend*
crash.

Killing postmaster itself as you are doing does not invoke the feature.

Try choosing any client backend by PID and then do your kill -9 on that
one.

Then expect any other backends to be killed also but the postmaster
still working and then spawning more backend when clients attempt to
connect.

>
> Regards
> Wolfgang

-- 
Jerry Sievers
Postgres DBA/Development Consulting
e: postgres.consulting@comcast.net
p: 312.241.7800