On Fri, 2020-10-09 at 05:44 +0200, Guillaume Lelarge wrote:
> Le ven. 9 oct. 2020 à 05:33, <p.sun.fun@gmail.com> a écrit :
> > On Thu, 2020-10-08 at 16:56 -0400, Tom Lane wrote:
> > > Rob Sargent <robjsargent@gmail.com> writes:
> > > > OK, well that’s a special db. Didn’t know it was that special,
> > > > though!
> > >
> > > It's not that special. The issue here is that each session is
> > > connecting
> > > to template1 and then trying to clone template1. You can't clone
> > an
> > > active database, because you might not get a consistent copy.
> > CREATE
> > > DATABASE knows that its own session isn't concurrently making any
> > > changes, so it allows copying the current database --- but it
> > can't
> > > know what some other session is doing, so if it sees some other
> > > session
> > > is also connected to the source database, it spits up.
> > >
> > > As I already said, routinely connecting to template1 is pretty
> > bad
> > > practice to start with, so the preferred answer is "don't do
> > that".
> > >
> > > regards, tom lane
> >
> > Thank you, guys. I will switch to the "postgres" database as a
> > default
> > one. IMHO, it is worth adding to the documentation into the CREATE
> > DATABASE section. I am glad that PostgreSQL has a strong community
> > that
> > stays behind the product.
>
> It's already in the documentation:
>
> "Although it is possible to copy a database other than template1 by
> specifying its name as the template, this is not (yet) intended as a
> general-purpose “COPY DATABASE” facility. The principal limitation is
> that no other sessions can be connected to the template database
> while it is being copied. CREATE DATABASE will fail if any other
> connection exists when it starts; otherwise, new connections to the
> template database are locked out until CREATE DATABASE completes.
> See Section 22.3 for more information."
>
> See https://www.postgresql.org/docs/13/sql-createdatabase.html.
Yep, you are right. Probably did't read carefully. Thanks for pointing
out.