Subject: Re: crashes due to setting max_parallel_workers=0
To: David Rowley <david.rowley@2ndquadrant.com>,
 Rushabh Lathia <rushabh.lathia@gmail.com>
References: <6ecd6f17-0dcf-1de7-ded8-0de7db1ddc88@2ndquadrant.com>
 <CAKJS1f-jnPUKyeF5OGJPNDtkkqMGUOcK+Yn6qjPzP4bTU04WXg@mail.gmail.com>
 <CAGPqQf2C6W8aNy9v9ZW2fJs+7p0MoBFfHEtXG8yypimGXTRUvw@mail.gmail.com>
 <CAKJS1f8x1ptsCidf2FpRXySUJfWLEMjSwxV8_kvozHFPQYaAow@mail.gmail.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>,
 Robert Haas <robertmhaas@gmail.com>
From: Tomas Vondra <tomas.vondra@2ndquadrant.com>
Message-ID: <eb41cd2f-efb7-698d-f6ef-e08e58c5693f@2ndquadrant.com>
Date: Mon, 27 Mar 2017 01:23:12 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
 Thunderbird/45.4.0
MIME-Version: 1.0
In-Reply-To: 
 <CAKJS1f8x1ptsCidf2FpRXySUJfWLEMjSwxV8_kvozHFPQYaAow@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

On 03/25/2017 02:01 PM, David Rowley wrote:
 >
> I wondered if there's anything we can do here to better test cases
> when no workers are able to try to ensure the parallel nodes work
> correctly, but the more I think about it, the more I don't see wrong
> with just using SET max_parallel_workers = 0;
>

It's demonstrably a valid way to disable parallel queries (i.e. there's 
nothing wrong with it), because the docs say this:

    Setting this value to 0 disables parallel query execution.

 >
> My vote would be to leave the GUC behaviour as is and add some tests
> to select_parallel.sql which exploit setting max_parallel_workers to 0
> for running some tests.
>
> If that's not going to fly, then unless we add something else to allow
> us to reliably not get any workers, then we're closing to close the
> door on being able to write automatic tests to capture this sort of
> bug.
>
> ... thinks for a bit....
>
> perhaps some magic value like -1 could be used for this... unsure of
> how that would be documented though...
>

I agree it'd be very useful to have a more where we generate parallel 
plans but then prohibit starting any workers. That would detect this and 
similar issues, I think.

I'm not sure we need to invent a new magic value, though. Can we simply 
look at force_parallel_mode, and if it's 'regress' then tread 0 differently?

regards

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


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers