Date: Fri, 6 Sep 2019 10:27:02 -0400
From: Alvaro Herrera from 2ndQuadrant <alvherre@alvh.no-ip.org>
To: Michael Paquier <michael@paquier.xyz>
Cc: Robert Haas <robertmhaas@gmail.com>,
	Masahiko Sawada <sawada.mshk@gmail.com>,
	Tatsuro Yamada <tatsuro.yamada.tf@nttcom.co.jp>,
	Rafia Sabih <rafia.pghackers@gmail.com>,
	Dmitry Dolgov <9erthalion6@gmail.com>, Peter Geoghegan <pg@bowt.ie>,
	PostgreSQL-development <pgsql-hackers@postgresql.org>,
	Jeff Janes <jeff.janes@gmail.com>,
	Thomas Munro <thomas.munro@enterprisedb.com>,
	Michael Paquier <michael.paquier@gmail.com>,
	Peter Eisentraut <peter.eisentraut@2ndquadrant.com>
Subject: Re: [HACKERS] CLUSTER command progress monitor
Message-ID: <20190906142702.GA8947@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20190906054418.GC1608@paquier.xyz>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Sep-06, Michael Paquier wrote:

> Finally comes the question of what do we do for v12?  I am adding in
> CC Peter, Alvaro being already present, who have been involved in the
> commits with CREATE INDEX and REINDEX.  It would be sad to revert a
> this feature, but well I'd rather do that now than regret later
> releasing the feature as it is currently shaped..  Let's see what the
> others think.

As far as I understand, CREATE INDEX is not affected -- only REINDEX is.
Of course, it would be sad to revert even the latter, but it's not as
bleak as reverting the whole thing.

That said, I did spend some time on this type of issue when doing CREATE
INDEX support; you can tell because I defined the columns for block
numbers in a scan separately from CREATE INDEX specific fields,
precisely to avoid multiple commands running concurrently from
clobbering unrelated columns:

/* Block numbers in a generic relation scan */
#define PROGRESS_SCAN_BLOCKS_TOTAL				15
#define PROGRESS_SCAN_BLOCKS_DONE				16

I would say that it's fairly useful to have CLUSTER report progress on
indexes being created underneath, but I understand that it might be too
late to be designing the CLUSTER report to take advantage of the CREATE
INDEX metrics.

I think a workable, not terribly invasive approach is to have REINDEX
process its commands conditionally: have the caller indicate whether
progress is to be reported, and skip the calls if not.  That would
(should) prevent it from clobbering the state set up by CLUSTER.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services