Re: [PATCH] Stop ALTER SYSTEM from making bad assumptions

Greetings,

On Sat, Jun 22, 2019 at 17:43 Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:

> On Sun, Jun 23, 2019 at 2:43 AM Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> >
> > Greetings,
> >
> > On Sat, Jun 22, 2019 at 17:07 Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
> wrote:
> >>
> >> On Fri, Jun 21, 2019 at 8:15 PM Robert Haas <robertmhaas(at)gmail(dot)com>
> wrote:
> >> >
> >> > On Mon, Jun 17, 2019 at 10:50 AM Ian Barwick
> >> > <ian(dot)barwick(at)2ndquadrant(dot)com> wrote:
> >> > > In Pg12, the code in pg_basebackup implies the correct thing to do
> is append to .auto.conf,
> >> > > but as demonstrated that can cause problems with duplicate entries.
> >> >
> >> > Yeah.
> >> >
> >> > To me, forcing every tools author to use postgresql.conf parsing tools
> >> > rather than just appending to the file is a needless burden on tool
> >> > authors.
> >> >
> >>
> >> OTOH, if we give license to all the tools that they can append to the
> >> .auto.conf file whenever they want, then, I think the contents of the
> >> file can be unpredictable. Basically, as of now, we allow only one
> >> backend to write to the file, but giving a free pass to everyone can
> >> create a problem. This won't be a problem for pg_basebackup, but can
> >> be for other tools.
> >
> >
> > I don’t think anyone was suggesting that tools be allowed to modify the
> file while the server is running- if a change needs to be made while the
> server is running, then it should be done through a call to ALTER SYSTEM.
> >
> > There’s no shortage of tools that, particularly with the merger in of
> recovery.conf, want to modify and manipulate the file when the server is
> down though.
> >
> > All that said, whatever code it is that we write for pg_basebackup to do
> this properly should go into our client side library, so other tools can
> leverage that and avoid having to write it themselves.
> >
>
> Fair enough. In that case, don't we need some mechanism to ensure
> that if the API fails, then the old contents are retained? Alter
> system ensures that by writing first the contents to a temporary file,
> but I am not sure if whatever is done by pg_basebackup has that
> guarantee.

I’m not sure that’s really the same. Certainly, pg_basebackup needs to
deal with a partial write, or failure of any kind, in a clean way that
indicates the backup isn’t good. The important bit is that the resulting
file be one that ALTER SYSTEM and potentially other tools will be able to
work with.

Thanks,

Stephen

>