From: Tom Lane <tgl@sss.pgh.pa.us>
To: =?UTF-8?Q?Ulf_Lohbr=C3=BCgge?= <ulf.lohbruegge@gmail.com>
cc: pgsql-performance@postgresql.org
Subject: Re: High CPU Usage of "SET ROLE"
In-reply-to: 
 <CABZYQRLv5TXS9P5VoK4n0LUy2KK0fmLPFCiCc5YHoGH5oSANyA@mail.gmail.com>
References: 
 <CABZYQRJDKytuyyL-th+Kq2_D4fS-rrb3sgDG2Y4qsNT__GGD8Q@mail.gmail.com>
 <82967.1540216668@sss.pgh.pa.us>
 <CABZYQRLv5TXS9P5VoK4n0LUy2KK0fmLPFCiCc5YHoGH5oSANyA@mail.gmail.com>
Comments: In-reply-to =?UTF-8?Q?Ulf_Lohbr=C3=BCgge?=
 <ulf.lohbruegge@gmail.com>
	message dated "Tue, 30 Oct 2018 20:49:32 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <17062.1540931263.1@sss.pgh.pa.us>
Date: Tue, 30 Oct 2018 16:27:43 -0400
Message-ID: <17063.1540931263@sss.pgh.pa.us>
Precedence: bulk

=?UTF-8?Q?Ulf_Lohbr=C3=BCgge?= <ulf.lohbruegge@gmail.com> writes:
> I think I have found something here. It looks like that the order of
> statements is affecting their duration. I somehow have the feeling that the
> first statement after "RESET ROLE;" experiences a performance degradation.

Hm.  It's well known that the first query executed in a *session* takes
a pretty big performance hit, because of the need to populate the
backend's catalog caches.  I'm not very sure however why "RESET ROLE"
would result in a mass cache flush, if indeed that's what's happening.

			regards, tom lane