Suboptimal plan choice problem with 8.3RC2 - Mailing list pgsql-hackers

From Guillaume Smet
Subject Suboptimal plan choice problem with 8.3RC2
Date
Msg-id 1d4e0c10801221052j3b530ae8n33f7bfc359b1cfd5@mail.gmail.com
Whole thread Raw
Responses Re: Suboptimal plan choice problem with 8.3RC2
List pgsql-hackers
Hi -hackers,

While testing RC2 on the new servers of one of our customers, I found
a query really slow on the new server which is quite fast on the old
box currently in production (production is 8.1 at the moment).
If I set enable_nestloop to off, the query is fast (the plan is
different from the 8.1 plan though).

I attached:
- the plan with regular configuration;
- the plan after disabling nested loops;
- the plan obtained with 8.1 on the current production box;
- the relevant configuration and schema of the concerned tables.

The plans are really too different to find a narrower test case so
it's the real test case.

Feel free to ask any additional information or tests.

Regards,

--
Guillaume

Attachment

pgsql-hackers by date:

Previous
From:
Date:
Subject: Re: Polyphase Merge
Next
From: Neil Conway
Date:
Subject: Re: autonomous transactions