Date: Wed, 23 Feb 2022 20:00:05 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: ldh@laurent-hasson.com
Cc: pgsql-performance@postgresql.org
Subject: Re: An I/O error occurred while sending to the backend (PG 13.4)
Message-ID: <20220224020005.GT9008@telsasoft.com>
References: 
 <MN2PR15MB2560DC71C217FF03AC9082F5856B9@MN2PR15MB2560.namprd15.prod.outlook.com>
 <20211204175926.GC17618@telsasoft.com>
 <MN2PR15MB256038B438B8990D08B4A991856B9@MN2PR15MB2560.namprd15.prod.outlook.com>
 <MN2PR15MB256034941F279136BA555236853D9@MN2PR15MB2560.namprd15.prod.outlook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <MN2PR15MB256034941F279136BA555236853D9@MN2PR15MB2560.namprd15.prod.outlook.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20220224020005.GT9008%40telsasoft.com>
Precedence: bulk

You originally mailed about an error on the client, and now you found
corresponding server logs, which suggests a veritable network issue.

Are the postgres clients and server on the same subnet ?  If not, what are the
intermediate routers ?  Is there any NAT happening ?  Do those devices have any
interesting logs that correspond with the server/client connection failures ?

Have you tried enabling TCP keepalives ?  This might help to convince a NAT
device not to forget about your connection.

https://www.postgresql.org/docs/current/runtime-config-connection.html
tcp_keepalives_idle=9
tcp_keepalives_interval=9
tcp_keepalives_count=0
tcp_user_timeout=0 -- You apparently have this set, but it cannot work on windows, so just generates noise.

On linux, you can check the keepalive counters in "netstat -not" to be sure
that it's enabled.  A similar switch hopefully exists for windows.

-- 
Justin