Date: Thu, 4 May 2017 10:54:49 -0300
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-hackers@postgreSQL.org
Subject: Re: Reducing runtime of stats regression test
Message-ID: <20170504135449.zepyxepq6ug5pgim@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <17795.1493869423@sss.pgh.pa.us>
User-Agent: NeoMutt/20170306-137-4415bd-dirty (1.8.0)
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

Tom Lane wrote:

> The other significant delay in stats.sql is
> 
> -- force the rate-limiting logic in pgstat_report_stat() to time out
> -- and send a message
> SELECT pg_sleep(1.0);
> 
> Now, we do seem to need a delay there, because the rate-limiting logic
> is unlikely to have permitted the count from the immediately preceding
> statement to have gotten sent right then, and the count won't get
> sent at all while we're inside wait_for_stats (since backends only
> send stats just before going idle).  But there's more than one way
> to skin this cat.  We can just start a new connection with \c, and
> let wait_for_stats wait for the old one to send its stats before quitting.
> Even on my oldest and slowest buildfarm machines, starting a new session
> takes well under one second.

So you changed table prevstats from temp to permanent; perhaps make it
unlogged?

I wonder if it'd be useful to have a "pg_stat_flush" or something, which
sends out whatever is queued in this session.  Then you wouldn't need
the reconnection.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers