Date: Sun, 29 Sep 2019 12:27:09 -0300
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: James Coleman <jtc331@gmail.com>
Cc: Bruce Momjian <bruce@momjian.us>,
	pgsql-hackers <pgsql-hackers@postgresql.org>
Subject: Re: [DOC] Document concurrent index builds waiting on each other
Message-ID: <20190929152709.GA25027@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CAAaqYe9P95wxtoOaF0KozJsVMZ4Dxv8982x-uKAur8Umb1GkEg@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Sep-28, James Coleman wrote:

> I believe caveats like this are worth calling out rather than
> expecting users to have to understand the implementation details an
> work out the implications on their own.

I agree.

> I read Alvaro as referring to the fact that the docs already call out
> the following:
> 
> > Regular index builds permit other regular index builds on the same
> > table to occur simultaneously, but only one concurrent index build
> > can occur on a table at a time.

Yeah, that's what I was understanding.

BTW I think there's an approach that could alleviate part of this
problem, at least some of the time: whenever CIC runs for an index
that's not on expression and not partial, we could set the
PROC_IN_VACUUM flag.  That would cause it to get ignored by other
processes for snapshot purposes (including CIC itself), as well as by
vacuum.  I need to take some time to research the safety of this, but
intuitively it seems safe.

Even further, I think we could also do it for regular CREATE INDEX
(under the same conditions) provided that it's not run in a transaction
block.  But that requires even more research/proof.

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