From: Tom Lane <tgl@sss.pgh.pa.us>
To: "Peter J. Holzer" <hjp-pgsql@hjp.at>
cc: pgsql-general@postgresql.org
Subject: Re: expected authentication request from server, but received H
In-reply-to: <20240523070649.7j6r2lidqrmhmr3o@hjp.at>
References: <20240523070649.7j6r2lidqrmhmr3o@hjp.at>
Comments: In-reply-to "Peter J. Holzer" <hjp-pgsql@hjp.at>
	message dated "Thu, 23 May 2024 09:06:49 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <524369.1716499394.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 23 May 2024 17:23:14 -0400
Message-ID: <524370.1716499394@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/524370.1716499394%40sss.pgh.pa.us>
Precedence: bulk

"Peter J. Holzer" <hjp-pgsql@hjp.at> writes:
> One of our users reports getting the error message
> "expected authentication request from server, but received H"
> when trying to connect to the database.

That's very bizarre, and I don't recall any similar reports in the
recent past.

> Server is PostgreSQL 15.7 (Ubuntu 15.7-1.pgdg22.04+1)
> I haven't been able to find out what client version they are using yet,
> but I suspect it's quite old.

It would definitely be interesting to find out what the client-side
code is, including the openssl version being used.

> We've been on 15.x for quite a while now and it did work until at least
> a month ago, so I wouldn't expect an incompatible protocol change,
> however, I think the upgrade from 15.6 to 15.7 was just last week, so
> maybe something changed between these minor versions?

A look at the release notes doesn't suggest that anything was changed
in that area.  But I wonder if your new build could differ in some
other relevant way, such as being linked to a newer openssl version.

> I'm especially intrigued by "received H". According to
> https://www.postgresql.org/docs/current/protocol-message-formats.html
> Byte1('H') could mark a Copy Out response or a Flush command. Both don't
> make sense in that context.

Yeah, that couldn't be a valid message type byte here.  I guess that
either (a) the TLS stuff is going wrong and delivering garbage
data, or (b) libpq is somehow out of sync with the server, such that
it's taking a data byte as a message type byte.

I found a description of an ancient instance of (b) here:

https://www.postgresql.org/message-id/flat/CAAZKuFaxdPccCs9%2B2hTbMM5iwPZP=
6C494sAVz6qZhOPcYSGkTQ%40mail.gmail.com

Of course we fixed that a dozen years ago, and anyway that particular
problem required a server that is in dire straits.

			regards, tom lane