Re: planner with index scan cost way off actual cost,

From: Simon Riggs <simon(at)2ndquadrant(dot)com>
To: Guillaume Cottenceau <gc(at)mnc(dot)ch>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: planner with index scan cost way off actual cost,
Date: 2006-03-21 20:57:27
Message-ID: 1142974647.24487.518.camel@localhost.localdomain
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Fri, 2006-03-17 at 11:09 +0100, Guillaume Cottenceau wrote:

> INFO: index "idx_sent_msgs_date_theme_status" now contains 3692284 row versions in 88057 pages

> SET effective_cache_size = 10000;

SET effective_cache_size > 88057, round up to 100000

to ensure the index cost calculation knows the whole index will be
cached, which it clearly could be with 4GB RAM.

If the cost is still wrong, it is because the index order doesn't
correlate physically with the key columns. Use CLUSTER.

Best Regards, Simon Riggs

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Scott Marlowe 2006-03-21 21:19:22 Re: Migration study, step 1: bulk write performance
Previous Message Simon Riggs 2006-03-21 20:44:50 Re: Migration study, step 1: bulk write performance