Horton

Re: Emit fewer vacuum records by reaping removable tuples during pruning

Melanie Plageman <melanieplageman@gmail.com>

View thread View on postgr.es
From: Melanie Plageman <melanieplageman@gmail.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: Peter Geoghegan <pg@bowt.ie>, Pg Hackers <pgsql-hackers@postgresql.org>, Andres Freund <andres@anarazel.de>, Robert Haas <robertmhaas@gmail.com>
Date: 2023-12-27T16:26:52Z
Lists: pgsql-hackers 
On Sat, Dec 23, 2023 at 9:14 PM Michael Paquier <michael@paquier.xyz> wrote:
>
> On Mon, Nov 13, 2023 at 07:06:15PM -0500, Melanie Plageman wrote:
> > As best I can tell, our best case scenario is Thomas' streaming read API
> > goes in, we add vacuum as a user, and we can likely remove the skip
> > range logic.
>
> This does not prevent the work you've been doing in 0001 and 0002
> posted upthread, right? Some progress is always better than no
> progress

Correct. Peter and I were mainly discussing next refactoring steps as
we move toward combining the prune, freeze, and VM records. This
thread's patches stand alone.

> I can see the appeal behind doing 0001 actually to keep
> the updates of the block numbers closer to where we determine if
> relation truncation is safe of not rather than use an intermediate
> state in LVPagePruneState.

Exactly.

> 0002 is much, much, much trickier..

Do you have specific concerns about its correctness? I understand it
is an area where we have to be sure we are correct. But, to be fair,
that is true of all the pruning and vacuuming code.

- Melanie

Commits

  1. Combine FSM updates for prune and no-prune cases.

  2. Remove LVPagePruneState.

  3. Move VM update code from lazy_scan_heap() to lazy_scan_prune().

  4. Optimize vacuuming of relations with no indexes.

  5. Be more consistent about whether to update the FSM while vacuuming.

  6. Remove hastup from LVPagePruneState.

  7. Use scanned_pages to decide when to failsafe check.

  8. Simplify lazy_scan_heap's handling of scanned pages.

  9. While vacuuming a large table, update upper-level FSM data every so often.