Message-ID: <4EC2CDE2.3000400@2ndQuadrant.com>
Date: Tue, 15 Nov 2011 15:38:58 -0500
From: Greg Smith <greg@2ndQuadrant.com>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
	rv:1.9.1.16) Gecko/20111110 Icedove/3.0.11
MIME-Version: 1.0
To: pgsql-admin@postgresql.org
Subject: Re: cancelled queries on standby
References: <1321354003378-4994025.post@n5.nabble.com>
In-Reply-To: <1321354003378-4994025.post@n5.nabble.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 11/15/2011 05:46 AM, MirrorX wrote:
> i am facing an issue concerning cancelled queries made on a hot standby
> server. if we suppose that there are some tables that are never changed (no
> inserts/updates/deletes) and that autovacuum is set to false for these
> tables, what else could there be affecting these tables and consequently
> cancelling my queries? thx in advance for any suggestions
>    

Query cancellation isn't targeted this way.  Whether or not a query 
needs to be canceled by vacuum cleanup depends on whether any open 
transaction can see the rows that are being cleaned up.  I think you 
have the understandable belief what rows are actually being viewed by 
the query running on the standby matter; they don't.

Let's say you have a long-running query Q accessing a table A on the 
standby.  Then some vacuum cleanup of table B happens on the master.  Q 
can then be canceled when those cleanup records are replicated.  The 
fact that A is never changed and doesn't have autovacuum enabled doesn't 
matter.  Because it's possible for Q to see B, it's canceled regardless 
of whether it's looked at it yet.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us