Date: Thu, 09 Feb 2023 11:31:49 +0900 (JST)
Message-Id: <20230209.113149.2028095933259376311.horikyota.ntt@gmail.com>
To: kuroda.hayato@fujitsu.com
Cc: katsuragiy@oss.nttdata.com, yuzhihong@gmail.com, tgl@sss.pgh.pa.us,
 vignesh21@gmail.com, pgsql-hackers@lists.postgresql.org,
 onderkalaci@gmail.com, masao.fujii@oss.nttdata.com,
 Shinya11.Kato@oss.nttdata.com
Subject: Re: [Proposal] Add foreign-server health checks infrastructure
From: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
In-Reply-To: <TYAPR01MB586675882721E646693E13FBF5CC9@TYAPR01MB5866.jpnprd01.prod.outlook.com>
References: <TYAPR01MB5866969DBAEEEC5AE07591BFF5CE9@TYAPR01MB5866.jpnprd01.prod.outlook.com>
	<TYAPR01MB5866B55A252EF7005AA81D5BF5CC9@TYAPR01MB5866.jpnprd01.prod.outlook.com>
	<TYAPR01MB586675882721E646693E13FBF5CC9@TYAPR01MB5866.jpnprd01.prod.outlook.com>
User-Agent: Mew version 6.8 on Emacs 26.1
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/20230209.113149.2028095933259376311.horikyota.ntt%40gmail.com>
Precedence: bulk

At Fri, 27 Jan 2023 06:57:01 +0000, "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com> wrote in 
> I found cfbot failure, PSA fixed version.

+       Unlike <xref linkend="libpq-PQstatus"/>, this function checks socket
+       health. This check is performed by polling the socket. This function is
+       currently available only on systems that support the non-standard
+       <symbol>POLLRDHUP</symbol> extension to the <symbol>poll</symbol> system

I find it quite confusing that we have pqSocketCheck and PQconnCheck,
that does almost the same thing.. Since pqSocketCheck is a static
function, we can modify the function as we like.

I still don't understand why we need pqconnCheck_internal separate
from pqSocketPoll(), and PQconnCheck from pqSocketCheck.

https://www.postgresql.org/message-id/flat/TYAPR01MB58665BF23D38EDF10028DE2AF5299%40TYAPR01MB5866.jpnprd01.prod.outlook.com#47d21431bf9fa94f763c824f6e81fa54
> IIUC, pqSocketCheck () calls pqSocketPoll(),
> and in the pqSocketPoll() we poll()'d the POLLIN or POLLOUT event.
> But according to [1], we must wait POLLRDHUP event,
> so we cannot reuse it straightforward.

Yeah, I didn't suggest to use the function as-is. Couldn't we extend
the fucntion by letting it accept end_time = 0 && !forRead &&
!forWrite, not causing side effects?

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center