Feedback-ID: ia2694551:Fastmail
Date: Fri, 24 Mar 2023 10:35:47 +0100
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Greg Stark <stark@mit.edu>
Cc: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>,
	Euler Taveira <euler@eulerto.com>,
	Amit Kapila <amit.kapila16@gmail.com>,
	Kyotaro Horiguchi <horikyota.ntt@gmail.com>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: add log messages when replication slots become active and
 inactive (was Re: Is it worth adding ReplicationSlot active_pid to
 ReplicationSlotPersistentData?)
Message-ID: <20230324093547.amsbfoql2kmbqco2@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CAM-w4HORsLbxwv8KVUHWu1PxKQLuCN1JjAnXknP8pFNSQfbh3A@mail.gmail.com>
Archived-At: <https://www.postgresql.org/message-id/20230324093547.amsbfoql2kmbqco2%40alvherre.pgsql>
Precedence: bulk

On 2023-Mar-23, Greg Stark wrote:

> On Thu, 23 Mar 2023 at 23:30, Bharath Rupireddy
> <bharath.rupireddyforpostgres@gmail.com> wrote:
> >
> > > + ereport(log_replication_commands ? LOG : DEBUG3,
> > > + (errmsg("acquired physical replication slot \"%s\"",
> > > + slotname)));
> 
> So this is just a bit of bike-shedding but I don't feel like these log
> messages really meet the standard we set for our logging. Like what
> did the acquiring? What does "acquired" actually mean for a
> replication slot? Is there not any meta information about the
> acquisition that can give more context to the reader to make this
> message more meaningful?
> 
> I would expect a log message like this to say, I dunno, something like
> "physical replication slot \"%s\" acquired by streaming TCP connection
> to 192.168.0.1:999 at LSN ... with xxxMB of logs to read"

Hmm, I don't disagree with your argument in principle, but I think this
proposal is going too far.  I think stating the PID is more than
sufficient.  And I don't think we need this patch to go great lengths to
explain what acquisition is, either; I mean, maybe that's a good thing
to have, but then that's a different patch.

> I even would be wondering if the other end shouldn't also be logging a
> corresponding log and we shouldn't be going out of our way to ensure
> there's enough information to match them up and presenting them in a
> way that makes that easy.

Hmm, you should be able to match things using the connection
information.  I don't think the slot acquisition operation in itself is
that important.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/
"Use it up, wear it out, make it do, or do without"