Date: Wed, 26 Mar 2008 12:49:56 -0800
From: Vinubalaji Gopal <vgopal@abaca.com>
To: "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: vacuum in Postgresql 8.0.x slowing down the database
Message-ID: <20080326124956.49e1c13d@vinu-ubuntu.abaca>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit

Hey all,
   I had posted sometime back asking for the best way to perform vacuum
with a lower priority - I did tune it up to a lower priority
and still noticed that the other database queries are slowing down
with a vacuum on one big table. I also tried to upgrade Postgresql to
8.0.15 as suggested and I still could reproduce the problem.

It happens when I try to vacuum one particular table which has 9.5
million rows and all the other inserts/selects are slowing down by a
factor of 10 times. I am using
vacuum_cost_delay = 100
vacuum_cost_limit = 200

Even if I cancel the vacuum operation or let the vacuum complete - the
slowdown continues to persist till I restart my application.

Whats the best way to analyze the reason Postgresql is slowing down? I
had a look at pg_locks (did not know what to look for) and also tried
starting postgresql with the debug option using: postmaster -d 5
-D /var/pgdata (saw a lot of output including the query being performed
but could not gather anything useful) 

The big table has never been reindexed and has a primary, unique key
with btree index and one foreign key constraint.

--
Vinu