Re: Log statement sample - take two - Mailing list pgsql-hackers

From Tomas Vondra
Subject Re: Log statement sample - take two
Date
Msg-id 20191104010807.e2tanu55ny2qqpez@development
Whole thread Raw
In response to Log statement sample - take two  (Adrien Nayrat <adrien.nayrat@anayrat.info>)
Responses Re: Log statement sample - take two  (Adrien Nayrat <adrien.nayrat@anayrat.info>)
List pgsql-hackers
On Sat, Oct 19, 2019 at 05:02:01PM +0200, Adrien Nayrat wrote:
>Hello,
>
>This patch propose a new way to sample statement to logs.
>
>As a reminder, this feature was committed in PG12[1] then reverted[2] after the
>proposition of log_statement_sample_limit[3]
>
>The first implementation added a new GUC to sample statement logged by
>log_min_duration_statement. Then, we wanted to add the ability to log all
>statement whose duration exceed log_statement_sample_limit.
>
>This was confusing because log_min_duration behaves as a minimum to enable
>sampling. While log_statement_sample_limit behave as maximum to disable it.[4]
>
>Tomas Vondra proposed to use two minimum thresholds:
>
>> 1) log_min_duration_sample - enables sampling of commands, using the
>> existing GUC log_statement_sample_rate
>>
>> 2) log_min_duration_statement - logs all commands exceeding this
>
>This patch implement this idea.
>
>PS: I notice I forgot to mention "Only superusers can change this setting" in
>the log_transaction_sample_rate documentation. It attached a second patch to fix
>this.
>

Seems fine to me, mostly. I think the docs should explain how
log_min_duration_statement interacts with log_min_duration_sample.
Attached is a patch doing that, by adding one para to each GUC, along
with some minor rewordings. I think the docs are mixing "sampling"
vs. "logging" and "durations" vs. "statements" not sure.

I also think the two new sampling GUCs (log_min_duration_sample and
log_statement_sample_rate) should be next to each other. We're not
ordering the GUCs alphabetically anyway.

I plan to make those changes and push in a couple days.


regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Attachment

pgsql-hackers by date:

Previous
From: Thomas Munro
Date:
Subject: Re: let's make the list of reportable GUCs configurable (was Re: Add%r substitution for psql prompts to show recovery status)
Next
From: Thomas Munro
Date:
Subject: Re: fe-utils - share query cancellation code