Re: why there is not VACUUM FULL CONCURRENTLY?

Euler Taveira <euler(at)eulerto(dot)com> wrote:

> On Fri, Apr 4, 2025, at 1:38 PM, Antonin Houska wrote:
>
> Euler Taveira <euler(at)eulerto(dot)com> wrote:
>
> > +
> > + <warning>
> > + <para>
> > + The <command>FULL</command> parameter is deprecated in favor of
> > + <xref linkend="sql-repack"/>.
> > + </para>
> > + </warning>
> > +
> >
> > The warnings, notes, and tips are usually placed *after* the description.
>
> You probably mean the subsecions "Notes on Clustering" and "Notes on
> Resources". I moved them into the "Notes" section.
>
> No. I said that it should be put after the <para> not before.
>
> @@ -98,6 +98,14 @@ VACUUM [ ( <replaceable class="parameter">option</replaceable> [, ...] ) ] [ <re
> <varlistentry>
> <term><literal>FULL</literal></term>
> <listitem>
> +
> + <warning>
> + <para>
> + The <command>FULL</command> parameter is deprecated in favor of
> + <xref linkend="sql-repack"/>.
> + </para>
> + </warning>
> +
> <para>
> Selects <quote>full</quote> vacuum, which can reclaim more
> space, but takes much longer and exclusively locks the table.
>
> > + SELECT
> > + S.pid AS pid,
> > + S.datid AS datid,
> > + D.datname AS datname,
> > + S.relid AS relid,
> > + CASE S.param1 WHEN 1 THEN 'REPACK'
> > + END AS command,
> >
> > Do you really need command? IIUC REPACK is the only command that will used by
> > this view. There is no need to differentiate commands here.
>
> REPACK is a regular command, so why shouldn't it have its view? Just like
> CLUSTER has one (pg_stat_progress_cluster).
>
> You missed my point. IIRC the command is relevant in the
> pg_stat_progress_cluster because there are multiple commands (CLUSTER, VACUUM
> FULL). However, in this new view there will be only one command so it is not
> necessary to inform it.
>
> > + *
> > + * 'cmd' indicates which commands is being executed. REPACK should be the only
> > + * caller of this function in the future.
> >
> > command.
>
> Not sure I understand this comment.
>
> Singular form. ... which command is ...
>

This is the next version. It addresses these remaining concerns and also gets
rid of the unnecessary rules in gram.y (which complained about earlier).

--
Antonin Houska
Web: https://www.cybertec-postgresql.com