Message-ID: <3AEEDAF7.F1FA2AB6@intrinsity.com>
Date: Tue, 01 May 2001 10:49:11 -0500
From: John Coers <coers@intrinsity.com>
Reply-To: coers@intrinsity.com
MIME-Version: 1.0
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-general@postgresql.org
Subject: Re: Copy
References: <Pine.LNX.4.21.0104261634460.2565-100000@olympus.scw.org>
	<3AE89D2C.409BEF42@intrinsity.com>
	<3AEDB916.BE3123A0@intrinsity.com>
	<3AEDDECF.6B3ED075@intrinsity.com> <391.988668966@sss.pgh.pa.us>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Tom Lane wrote:
> 
> John Coers <coers@intrinsity.com> writes:
> > Does a COPY FROM block?  In case I am using the wrong terminology,
> > what I need to know is if I have multiple clients performing COPY FROM...PQputline()
> > using asynchronous connections, will I lose data?
> 
> No.
> 
> > Will the server simply execute them serially?
> 
> They'll be executed in parallel, just the same as if each client had
> done a bunch of INSERTs wrapped in a BEGIN/END block.  You'd only see
> problems if clients inserted conflicting data (eg, identical keys in
> a column with a UNIQUE index).  You may care to read the documentation
> about multi-version concurrency control and locking.
> 
>                         regards, tom lane

hmmmm...I get best performance with N=2 and force the clients to keep trying to
get a connection.  If I leave the default of N=32 the machine gets bogged down 
to a complete standstill by 32 copies of 15k rows of data.  The CPU doesn't seem 
stressed -- what may be causing the bottleneck?


-- 
John Coers            Intrinsity, Inc.        
coers@intrinsity.com  Austin, Texas