Horton

Re: PG 14 release notes, first draft

Matthias van de Meent <boekewurm+postgres@gmail.com>

View thread View on postgr.es
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
To: Bruce Momjian <bruce@momjian.us>
Cc: Matthias van de Meent <boekewurm+postgres@gmail.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2021-05-10T20:21:00Z
Lists: pgsql-hackers 
On Mon, 10 May 2021 at 19:34, Bruce Momjian <bruce@momjian.us> wrote:
>
> On Mon, May 10, 2021 at 01:44:12PM +0200, Matthias van de Meent wrote:
> > On Mon, 10 May 2021 at 08:03, Bruce Momjian <bruce@momjian.us> wrote:
> > >
> > > I have committed the first draft of the PG 14 release notes.  You can
> > > see the most current  build of them here:
> > >         https://momjian.us/pgsql_docs/release-14.html
> > >
> > > I need clarification on many items, and the document still needs its
> > > items properly ordered, and markup added.  I also expect a lot of
> > > feedback.
> >
> > I noticed that the improvement in bloat control in the HeapAM that I
> > know of (3c3b8a4b, 0ff8bbde) weren't documented here. Although each
> > can be considered minor, they together can decrease the bloating
> > behaviour of certain workloads significantly (and limit the total
> > damage), and in my opinion this should be mentioned.
> >
> > 3c3b8a4b: Returns space claimed for the line pointer array back to the
> > page's empty space, so that it can also be used for tuple data.
> >
> > 0ff8bbde: Allows large tuples to be inserted on pages which have only
> > a small amount of data, regardless of fillfactor.
> >
> > Together they should be able to help significantly in both bloat
> > prevention and bloat reduction.
>
> I looked at those items.  I try to mention performance items that enable
> new workloads or require some user action to benefit from it.

0ff8bbde Enables a workload that inserts (and non-locally updates)
large (> FILLFACTOR %) tuples in tables that have a low FILLFACTOR.
Previously this would fail dramatically by only inserting on new
pages; this would extend the table indefinately. See the thread [0]

3c3b8a4b improves workloads with high local update-then-delete churn.
Previously this would irreversably claim space on the page for tuple
identifiers even when they were later deleted; now we can reclaim this
space when a tuple is deleted from the page.

I see these two improvements in a similar light as the bottom-up index
deletion in btree: No user action required, works out-of-the-box,
decreases bloat / disk usage, but good to note as it fixes (known)
bloating footguns that a user might have encountered.

> I am not sure these two qualify, but can others comments?  Thanks.

I'd like to refer to Peter Geoghegan's reply [1] upthread.

Thank you for your effort,

Matthias van de Meent

[0] https://www.postgresql.org/message-id/flat/6e263217180649339720afe2176c50aa%40opammb0562.comp.optiver.com
[1] https://www.postgresql.org/message-id/CAH2-Wz%3D-A%3DjRxpB2Owj3KQadCue7%2BNLqj56Q566ees7TapMRvA%40mail.gmail.com

Commits

  1. doc: add mention of +4GB windows file handling in PG14 relnotes

  2. doc: PG 14 relnotes fixes

  3. doc: PG 14 relnote updates

  4. doc: PG 14 relnotes, adjust pg_{read|write}_all_data entry

  5. Update PG 14 relnotes for vacuum_cost_page_miss

  6. doc: update PG 14 release notes with recent feedback

  7. Update replication statistics after every stream/spill.

  8. Add information of total data processed to replication slot stats.

  9. Use NameData datatype for slotname in stats.

  10. Add back vacuum_cleanup_index_scale_factor parameter.

  11. Don't consider newly inserted tuples in nbtree VACUUM.

  12. pgbench: Improve time logic.

  13. Track total amounts of times spent writing and syncing WAL data to disk.

  14. Allow decoding at prepare time in ReorderBuffer.

  15. Extend the output plugin API to allow decoding of prepared xacts.

  16. Fix nbtree cleanup-only VACUUM stats inaccuracies.

  17. Execute invalidation messages for each XLOG_XACT_INVALIDATIONS message

  18. Fix our Windows stat() emulation to handle file sizes > 4GB.

  19. Trigger autovacuum based on number of INSERTs

  20. Speed up planning when partitions can be pruned at plan time.

  21. Remove support for password_encryption='off' / 'plain'.

  22. Increase work_mem and maintenance_work_mem defaults by 4x