To: Laszlo Nagy <gandalf@shopzeus.com>
cc: Alvaro Herrera <alvherre@commandprompt.com>, pgsql-admin@postgresql.org
Subject: Re: rebellious pg stats collector (reopened case) 
In-reply-to: <494F67D2.3090406@shopzeus.com> 
References: <49479876.8040607@shopzeus.com>
	<20081216134542.GH4741@alvh.no-ip.org>
	<494B614C.5060808@shopzeus.com>
	<20081219130446.GC4278@alvh.no-ip.org>
	<494F67D2.3090406@shopzeus.com>
Comments: In-reply-to Laszlo Nagy <gandalf@shopzeus.com>
	message dated "Mon, 22 Dec 2008 11:11:30 +0100"
Date: Mon, 22 Dec 2008 08:23:48 -0500
Message-ID: <6938.1229952228@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Laszlo Nagy <gandalf@shopzeus.com> writes:
> And the stats collector exited! It cannot be by accident. Last time I 
> checked, it had the same problem: tried to poll something. It was in an 
> inifinte loop, polling something but always got POLERR. Then because I 
> started to trace it, it exited with rval=0 or rval=1.

> I do not understand why it exited after I started to trace it? Is this 
> some kind of bug?

AFAICS, the only ways to reach the exit(0) in PgstatCollectorMain are
(1) need_exit becomes set (and the only way for that to happen is
to receive a SIGQUIT signal); or (2) PostmasterIsAlive(true) fails,
which implies that getppid() is returning something different than it
used to.

I wonder whether your tracing tool is affecting the result of
getppid().  Most people would consider that a bug in the tracing tool.

			regards, tom lane