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

another 8.1->8.4 regression

From: Ben Chobot <bench(at)silentmedia(dot)com>
To: PostgreSQL Performance <pgsql-performance(at)postgresql(dot)org>
Subject: another 8.1->8.4 regression
Date: 2010-02-16 21:29:16
Message-ID: 84633B63-C6F3-47A1-BC85-AE175BC43CCA@silentmedia.com (view raw or flat)
Thread:
Lists: pgsql-performance
I'm having problems with another one of my queries after moving from 8.1.19 to 8.4.2. On 8.1.19, the plan looked like this:

http://wood.silentmedia.com/bench/8119

That runs pretty well. On 8.4.2, the same query looks like this:

http://wood.silentmedia.com/bench/842_bad

If I turn off mergejoin and hashjoin, I can get 8.4.2 to spit out this:

http://wood.silentmedia.com/bench/842_better

...which it thinks is going to suck but which does not. 

The query and relevant table definitions are here:

http://wood.silentmedia.com/bench/query_and_definitions


Any suggestions? I'm guessing the problem is with the absurd over-estimation on the nested loop under the sort node, but I'm not sure why it's so bad. 

Responses

pgsql-performance by date

Next:From: Jeff DavisDate: 2010-02-16 21:34:11
Subject: Re: Linux I/O tuning: CFQ vs. deadline
Previous:From: lionel duboeufDate: 2010-02-16 16:47:41
Subject: Re: Almost infinite query -> Different Query Plan when changing where clause value

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