Re: COPY FREEZE and setting PD_ALL_VISIBLE/visibility map bits

Andres Freund <andres@anarazel.de>

From: Andres Freund <andres@anarazel.de>
To: Darafei "Komяpa" Praliaskouski <me@komzpa.net>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Alvaro Herrera <alvherre@2ndquadrant.com>, Pavan Deolasee <pavan.deolasee@gmail.com>, Masahiko Sawada <sawada.mshk@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>, Jeff Janes <jeff.janes@gmail.com>, Simon Riggs <simon@2ndquadrant.com>, Tomas Vondra <tomas.vondra@2ndquadrant.com>
Date: 2019-04-05T05:42:41Z
Lists: pgsql-hackers
Hi,

On 2019-04-05 08:38:34 +0300, Darafei "Komяpa" Praliaskouski wrote:
> On Fri, Apr 5, 2019 at 6:58 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> 
> > Andres Freund <andres@anarazel.de> writes:
> > > I think the right approach would be to do all of this in heap_insert and
> > > heap_multi_insert. Whenever starting to work on a page, if INSERT_FROZEN
> > > is specified, remember whether it is either currently empty, or is
> > > already marked as all-visible. If previously empty, mark it as all
> > > visible at the end. If already all visible, there's no need to change
> > > that. If not yet all-visible, no need to do anything, since it can't
> > > have been inserted with COPY FREEZE.  Do you see any problem doing it
> > > that way?
> >
> > Do we want to add overhead to these hot-spot routines for this purpose?
> >
> 
> Sizing the overhead: workflows right now don't end with COPY FREEZE - you
> need another VACUUM to set maps.
> Anything that lets you skip that VACUUM (and faster than that VACUUM
> itself) is helping. You specifically asked for it to be skippable with
> FREEZE anyway.

Tom's point was that the routines I was suggesting to adapt above aren't
just used for COPY FREEZE.

Greetings,

Andres Freund



Commits

  1. Fix COPY FREEZE with CLOBBER_CACHE_ALWAYS

  2. Set PD_ALL_VISIBLE and visibility map bits in COPY FREEZE

  3. docs: client certificates are always sent to the server

  4. Mark commit and abort WAL records with XLR_SPECIAL_REL_UPDATE.

  5. docs: land height is "elevation", not "altitude"

  6. Improve the internal implementation of ereport().