Feedback-ID: i0fe9450f:Fastmail
Date: Mon, 6 Oct 2025 12:17:46 +0900
From: Michael Paquier <michael@paquier.xyz>
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: "Aya Iwata (Fujitsu)" <iwata.aya@fujitsu.com>,
	pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: [PROPOSAL] Termination of Background Workers for ALTER/DROP
 DATABASE
Message-ID: <aOM02tVFtXmuR0M1@paquier.xyz>
References: 
 <OS7PR01MB11964335F36BE41021B62EAE8EAE4A@OS7PR01MB11964.jpnprd01.prod.outlook.com>
 <OSCPR01MB149662AEA64F4E66F494C2584F5E3A@OSCPR01MB14966.jpnprd01.prod.outlook.com>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="mSIJ+Llf+HQSShsf"
Content-Disposition: inline
In-Reply-To: 
 <OSCPR01MB149662AEA64F4E66F494C2584F5E3A@OSCPR01MB14966.jpnprd01.prod.outlook.com>
Archived-At: 
 <https://www.postgresql.org/message-id/aOM02tVFtXmuR0M1%40paquier.xyz>
Precedence: bulk


--mSIJ+Llf+HQSShsf
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Mon, Oct 06, 2025 at 01:59:08AM +0000, Hayato Kuroda (Fujitsu) wrote:
> Per my understanding, we already have a facility that terminates a backgr=
ound
> worker, TerminateBackgroundWorker(). So, I'm afraid your proposal has alr=
eady
> been done by combining this function and ProcessUtility_hook.

The main take that I am getting here from Iwata-san is that this would
lead to less code duplication.  Another item, which you are not
mentioning, is that this would be more flexible with bgworkers that
have been starting dynamically, where shared_preload_libraries may not
be used, still a bgworker would need to react.  So the suggestion of a
new API to control if a bgworker should be stopped like any other
backend when there is a database activity worth it is a good one, as
long as it is in line with what we do with normal backends.

AcceptBackgroundWorkerCancel() is going backwards, IMO.  Wouldn't it
be better to pass it down as an option in bgw_flags, to mark that a
bgworker connected to a database can be shutdown due to the effect of
a database getting dropped or moved?  There could be an argument
behind using bgw_extra, but that would not be in line with the
database connection argument which is a state defined when a bgworker
is registered.

> So, is the main benefit of the patch to shorten extensions codes which us=
es
> bgworker?

I'd ask for the addition of tests when it comes to such a facility,
and your proposal has none of that.  I would suggest worker_spi with
an option that can be passed to worker_spi_launch().=20
--
Michael

--mSIJ+Llf+HQSShsf
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----
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=Pgj5
-----END PGP SIGNATURE-----

--mSIJ+Llf+HQSShsf--