Skip site navigation (1) Skip section navigation (2)

Re: high user cpu, massive SELECTs, no io waiting problem

From: Greg Smith <greg(at)2ndquadrant(dot)com>
To: Thomas Pöhler <tp(at)turtle-entertainment(dot)de>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: high user cpu, massive SELECTs, no io waiting problem
Date: 2011-02-16 20:36:01
Message-ID: 4D5C3531.3070308@2ndquadrant.com (view raw or flat)
Thread:
Lists: pgsql-performance
Thomas Pöhler wrote:
> We are running a biweekly downtime where we do a complete reindex and vaccum full. We cannot identify certain queries causing this.

If you feel that you need VACUUM FULL, either something terribly wrong 
has happened, or someone has gotten confused.  In both cases it's 
unlikely you want to keep doing that.  See 
http://wiki.postgresql.org/wiki/VACUUM_FULL for a nice document leading 
through figuring what to do instead.

Note that if you have a database that fits in RAM, but is filled with 
the sort of index bloat garbage that using VACUUM FULL will leave 
behind, it will cause excessive CPU use when running queries.  If you 
already have planned downtime, you really should try to use use CLUSTER 
instead, to remove that from the list of possible causes for your issue.

-- 
Greg Smith   2ndQuadrant US    greg(at)2ndQuadrant(dot)com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us
"PostgreSQL 9.0 High Performance": http://www.2ndQuadrant.com/books


In response to

Responses

pgsql-performance by date

Next:From: Chris HooverDate: 2011-02-16 20:51:36
Subject: Estimating hot data size
Previous:From: Kevin GrittnerDate: 2011-02-16 19:22:26
Subject: Re: high user cpu, massive SELECTs, no io waiting problem

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group