Re: Function runtime increases after 5 calls in the same session.

From: Marti Raudsepp <marti(at)juffo(dot)org>
To: "Ilya I(dot) Ashchepkov" <koctep(at)gmail(dot)com>
Cc: pgsql <pgsql-general(at)postgresql(dot)org>
Subject: Re: Function runtime increases after 5 calls in the same session.
Date: 2014-10-21 10:20:28
Message-ID: CABRT9RDTp279tKSXTN0E8wH2UwJJKx9JGP-HTgrpHKFpBYid7g@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

Hi

On Tue, Oct 21, 2014 at 12:53 PM, Ilya I. Ashchepkov <koctep(at)gmail(dot)com> wrote:
> I wrote a function and during testing it I came across the strange
> behaviour.
> Function runtime is about 200ms first 5 times, 6th and futher calls takes
> ~22000 ms.
> I simplified my schema, you can see it in attached file.
> I've tested on 9.4beta3 and 9.3.5.

I didn't look at the test case. But most likely the problem is that
after 5 executions, plancache decides that a generic plan is no more
expensive than a custom plan, and reverts to executing the generic
one. But in reality there is a big difference in execution time.

See this for how the logic works:
https://github.com/postgres/postgres/blob/master/src/backend/utils/cache/plancache.c#L1036

As a workaround you could use PL/pgSQL EXECUTE to run the query, which
isn't subject to generic plans:
http://www.postgresql.org/docs/current/static/plpgsql-statements.html#PLPGSQL-STATEMENTS-EXECUTING-DYN

Regards,
Marti

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Michael Paquier 2014-10-21 13:45:55 Re: Yosemite (OSX 10.0) problems with Postgresql
Previous Message Craig Ringer 2014-10-21 10:04:26 Re: Yosemite (OSX 10.0) problems with Postgresql