Re: Implementing cost limit/delays for insert/delete/update/select

From: Simon Riggs <simon(at)2ndQuadrant(dot)com>
To: Peter Schuller <peter(dot)schuller(at)infidyne(dot)com>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: Implementing cost limit/delays for insert/delete/update/select
Date: 2008-08-26 11:03:37
Message-ID: 1219748617.5343.1310.camel@ebony.2ndQuadrant
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers


On Mon, 2008-08-25 at 22:39 +0200, Peter Schuller wrote:

> Does this sound vaguely sensible? Is there an obvious show-stopper I
> am missing?

This was a well structured proposal.

The main problem is where you put the delay_point() calls. If you put
them at the top of the executor then you will get a delay proportional
to the number of rows retrieved. For many queries, such as count(*) this
might be just one row, yet have run for hours. There's no point having a
priority scheme if it doesn't apply to all queries equally.

If you put them at each call of each node then you will get an
unacceptable overhead as Tom suggests.

Not sure what to suggest, if anything, apart from just writing your own
delay() function and using it in your query.

--
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Simon Riggs 2008-08-26 12:03:47 Re: gsoc, oprrest function for text search take 2
Previous Message Dave Cramer 2008-08-26 10:59:38 Re: can't stop autovacuum by HUP'ing the server