Horton

Re: BitmapHeapScan streaming read user and prelim refactoring

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Robert Haas <robertmhaas@gmail.com>
Cc: James Hunter <james.hunter.pg@gmail.com>, Melanie Plageman <melanieplageman@gmail.com>, Tomas Vondra <tomas@vondra.me>, Nazir Bilal Yavuz <byavuz81@gmail.com>, Dilip Kumar <dilipbalaut@gmail.com>, Heikki Linnakangas <hlinnaka@iki.fi>, Andres Freund <andres@anarazel.de>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2025-04-16T04:30:52Z
Lists: pgsql-hackers 
On Tue, Apr 15, 2025 at 5:44 AM Robert Haas <robertmhaas@gmail.com> wrote:
> On Thu, Apr 10, 2025 at 11:15 PM Thomas Munro <thomas.munro@gmail.com> wrote:
> > The new streaming BHS isn't just issuing probabilistic hints about
> > future access obtained from a second iterator.  It has just one shared
> > iterator connected up to the workers' ReadStreams.  Each worker pulls
> > a disjoint set of blocks out of its stream, possibly running a bunch
> > of IOs in the background as required.
>
> It feels to me like the problem here is that the shared iterator is
> connected to unshared read-streams. If you make a shared read-stream
> object and connect the shared iterator to that instead, does that
> solve this whole problem, or is there more to it?

More or less, yeah, just put the whole ReadStream object in shared
memory, pin an LWLock on it and call it a parallel-aware or shared
ReadStream.  But how do you make the locking not terrible?

My "work stealing" brain dump was imagining a way to achieve the same
net effect, except NOT have to acquire an exclusive lock for every
buffer you pull out of the stream.  I was speculating that we could
achieve zero locking for most of the stream without any cache line
ping pong, but a cunning read barrier scheme could detect when you've
been flipped into a slower coordination mode by another backend and
need to turn on some locking and fight over the last handful of
buffers.  And I was also observing that if you can figure out to make
it general and reusable enough, we have more unsolved problems like
this in unrelated parallel query code not even involving streams.
It's a tiny more approachable subset of the old "data buffered in
other workers" problem, as I think you called it once.

Commits

  1. Fix bitmapheapscan incorrect recheck of NULL tuples

  2. Increase default maintenance_io_concurrency to 16

  3. Separate TBM[Shared|Private]Iterator and TBMIterateResult

  4. Improve read_stream.c advice for dense streams.

  5. Increase default effective_io_concurrency to 16

  6. Delay extraction of TIDBitmap per page offsets

  7. Add lossy indicator to TBMIterateResult

  8. Move BitmapTableScan per-scan setup into a helper

  9. Add and use BitmapHeapScanDescData struct

  10. Fix bitmap table scan crash on iterator release

  11. Bitmap Table Scans use unified TBMIterator

  12. Add common interface for TBMIterators

  13. Make table_scan_bitmap_next_block() async-friendly

  14. Move EXPLAIN counter increment to heapam_scan_bitmap_next_block

  15. Refactor tidstore.c iterator buffering.

  16. BitmapHeapScan: Remove incorrect assert and reset field

  17. Change BitmapAdjustPrefetchIterator to accept BlockNumber

  18. BitmapHeapScan: Use correct recheck flag for skip_fetch

  19. BitmapHeapScan: Push skip_fetch optimization into table AM

  20. BitmapHeapScan: postpone setting can_skip_fetch

  21. BitmapHeapScan: begin scan after bitmap creation

  22. Fix EXPLAIN Bitmap heap scan to count pages with no visible tuples

  23. Remove redundant snapshot copying from parallel leader to workers

  24. Remove some obsolete smgrcloseall() calls.

  25. Remove the "snapshot too old" feature.

  26. Compute XID horizon for page level index vacuum on primary.