Re: auto_explain sample rate

From: Julien Rouhaud <julien(dot)rouhaud(at)dalibo(dot)com>
To: Magnus Hagander <magnus(at)hagander(dot)net>, Petr Jelinek <petr(at)2ndquadrant(dot)com>
Cc: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, Craig Ringer <craig(at)2ndquadrant(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: auto_explain sample rate
Date: 2016-03-11 13:51:00
Message-ID: 56E2CD44.1040203@dalibo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 11/03/2016 11:45, Magnus Hagander wrote:
>
> Coming back to the previous discussions about random() - AFAICT this
> patch will introduce the random() call always (in explain_ExecutorStart):
>
> +if (auto_explain_log_min_duration >= 0 && nesting_level == 0)
> +current_query_sampled = (random() < auto_explain_sample_ratio *
> +MAX_RANDOM_VALUE);
>
>
> Not sure what the overhead is, but wouldn't it be better to include a
> check for current_query_sampled>0 in the if part of that statement?
> Regardless of performance, that feels cleaner to me. Or am I missing
> something?

You mean check for auto_explain_sample_ratio > 0 ?

There would be a corner case if a query is sampled
(current_query_sampled is true) and then auto_explain_sample_ratio is
set to 0, all subsequent queries in this backend would be processed.

We could add a specific test for this case to spare a random() call, but
I fear it'd be overkill. Maybe it's better to document that the good way
to deactivate auto_explain is to set auto_explain.log_min_duration to -1.

--
Julien Rouhaud
http://dalibo.com - http://dalibo.org

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Robert Haas 2016-03-11 13:51:31 Re: dblink: add polymorphic functions.
Previous Message Robert Haas 2016-03-11 13:49:46 Re: [patch] Proposal for \crosstabview in psql