Re: Built-in connection pooler

From: Konstantin Knizhnik <knizhnik(at)garret(dot)ru>
To: Zhihong Yu <zyu(at)yugabyte(dot)com>
Cc: Daniel Gustafsson <daniel(at)yesql(dot)se>, Konstantin Knizhnik <k(dot)knizhnik(at)postgrespro(dot)ru>, Michael Paquier <michael(at)paquier(dot)xyz>, David Steele <david(at)pgmasters(dot)net>, "ideriha(dot)takeshi(at)fujitsu(dot)com" <ideriha(dot)takeshi(at)fujitsu(dot)com>, Jaime Casanova <jaime(dot)casanova(at)2ndquadrant(dot)com>, Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com>, Ryan Lambert <ryan(at)rustprooflabs(dot)com>, Thomas Munro <thomas(dot)munro(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>, Bruce Momjian <bruce(at)momjian(dot)us>, Dimitri Fontaine <dim(at)tapoueh(dot)org>, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>
Subject: Re: Built-in connection pooler
Date: 2021-03-22 09:05:53
Message-ID: 0d9c9165-3389-dd70-3429-df2facce44ec@garret.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,
Thank you for review!

On 21.03.2021 23:59, Zhihong Yu wrote:
> Hi,
>
> +          With <literal>load-balancing</literal> policy postmaster
> choose proxy with lowest load average.
> +          Load average of proxy is estimated by number of clients
> connection assigned to this proxy with extra weight for SSL connections.
>
> I think 'load-balanced' may be better than 'load-balancing'.

Sorry, I am not a native speaker.
But it seems to me (based on the articles I have read), then
"load-balancing" is more widely used term:

https://en.wikipedia.org/wiki/Load_balancing_(computing)

> postmaster choose proxy -> postmaster chooses proxy

Fixed.

>
> +          Load average of proxy is estimated by number of clients
> connection assigned to this proxy with extra weight for SSL connections.
>
> I wonder if there would be a mixture of connections with and without SSL.
Why not? And what is wrong with it?

>
> +         Terminate an idle connection pool worker after the specified
> number of milliseconds.
>
> Should the time unit be seconds ? It seems a worker would exist for at
> least a second.
>
Most of other similar timeouts: statement timeout, session timeout...
are specified in milliseconds.

> + * Portions Copyright (c) 1996-2018, PostgreSQL Global Development Group
>
> It would be better to update the year in the header.

Fixed.

>
> +        * Use then for launching pooler worker backends and report error
>
> Not sure I understand the above sentence. Did you mean 'them' instead
> of 'then' ?

Sorry, it is really mistyping.
"them" should be used.
Fixed.

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Andrei Zubkov 2021-03-22 09:06:15 Re: [PATCH] pg_stat_statements dealloc field ignores manual deallocation
Previous Message Peter Eisentraut 2021-03-22 09:00:39 Re: [PATCH] Identify LWLocks in tracepoints