Date: Mon, 5 Dec 2022 09:35:23 -0800
From: Nathan Bossart <nathandbossart@gmail.com>
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Subject: Re: suppressing useless wakeups in logical/worker.c
Message-ID: <20221205173523.GA2714833@nathanxps13>
References: <20221202195503.GD2277157@nathanxps13>
 <TYAPR01MB58669549D0137CC76A9BB6A0F5189@TYAPR01MB5866.jpnprd01.prod.outlook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <TYAPR01MB58669549D0137CC76A9BB6A0F5189@TYAPR01MB5866.jpnprd01.prod.outlook.com>
Archived-At: 
 <https://www.postgresql.org/message-id/20221205173523.GA2714833%40nathanxps13>
Precedence: bulk

On Mon, Dec 05, 2022 at 01:00:19PM +0000, Hayato Kuroda (Fujitsu) wrote:
> But in your patch, the apply worker calcurates wakeup[LRW_WAKEUP_PING] and
> wakeup[LRW_WAKEUP_TERMINATE] again when it gets SIGHUP, so the worker never sends
> ping with requestReply = true, and never exits due to the timeout.

This is the case for the walreceiver patch, too.  If a SIGHUP arrives just
before we are due to ping the server, the ping wakeup time will be pushed
back.  To me, this seems unlikely to cause any issues in practice unless
the server is being constantly SIGHUP'd.  If we wanted to fix this, we'd
probably need to recompute the wakeup times using the values currently set.
I haven't looked into this too closely, but it doesn't sound tremendously
difficult.  Thoughts?

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com