Horton
Menu

Re: eliminate xl_heap_visible to reduce WAL (and eventually set VM on-access)

Alexander Lakhin <exclusion@gmail.com>

View thread
From: Alexander Lakhin <exclusion@gmail.com>
To: Melanie Plageman <melanieplageman@gmail.com>, Andres Freund <andres@anarazel.de>
Cc: Tomas Vondra <tomas@vondra.me>, David Rowley <dgrowleyml@gmail.com>, Kirill Reshke <reshkekirill@gmail.com>, Chao Li <li.evan.chao@gmail.com>, Andrey Borodin <x4mmm@yandex-team.ru>, Xuneng Zhou <xunengzhou@gmail.com>, Robert Haas <robertmhaas@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Heikki Linnakangas <hlinnaka@iki.fi>
Date: 2026-04-20T19:00:00Z
Lists: pgsql-hackers 
Hello Melanie and Andres,

20.04.2026 19:18, Melanie Plageman wrote:
>> but I wonder if there are other queries,
>> which plans can change due to the same reason.
> I think we'll have to take this on a case-by-case basis when we see
> failures. While it is certainly possible other tests just rely on
> autovacuum not having run and set the page all-visible, many of them
> probably have already had to account for that.

Thank you for paying attention to this!

I think, I found another test which suffers from autoanalyze with the new
behavior: [1], [2].

Initially I reproduced this diff on a slow armv7 device after many
iterations of `make check` with:
autovacuum_naptime = 1
autovacuum_analyze_threshold = 1
debug_parallel_query = 'regress'

But now I see that it can be reproduced on an ordinary machine with just:
--- a/src/test/regress/sql/plancache.sql
+++ b/src/test/regress/sql/plancache.sql
@@ -208,2 +208,3 @@ execute test_mode_pp(1); -- 2x
  execute test_mode_pp(1); -- 3x
+analyze test_mode;
  execute test_mode_pp(1); -- 4x
(and expected/plancache.out updated)

and `make check` running in a loop. It failed for me on iterations 5, 4,
10 (as far as I can see, analyze updates relallvisible not every time):
# parallel group (18 tests):  prepare xml conversion plancache limit returning copy2 polymorphism sequence rowtypes 
largeobject temp rangefuncs with truncate domain plpgsql alter_table
# diff -U3 .../src/test/regress/expected/plancache.out .../src/test/regress/results/plancache.out
# --- .../src/test/regress/expected/plancache.out    2026-04-20 21:35:30.677775398 +0300
# +++ .../src/test/regress/results/plancache.out     2026-04-20 21:43:49.324492302 +0300
# @@ -374,11 +374,11 @@
#
#  -- we should now get a really bad plan
#  explain (costs off) execute test_mode_pp(2);
# -         QUERY PLAN
# ------------------------------
# +                        QUERY PLAN
# +----------------------------------------------------------
#   Aggregate
# -   ->  Seq Scan on test_mode
# -         Filter: (a = $1)
# +   ->  Index Only Scan using test_mode_a_idx on test_mode
# +         Index Cond: (a = $1)
#  (3 rows)
#
#  -- but we can force a custom plan

The same modified test survived 50 iterations at 378a21618~1.

[1] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dodo&dt=2026-04-07%2012%3A45%3A07
[2] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dodo&dt=2026-04-12%2022%3A45%3A06

Best regards,
Alexander

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Stabilize btree_gist test against on-access VM setting

  2. Stabilize plancache test against on-access VM setting

  3. Make local buffers pin limit more conservative

  4. Update FSM during prune/freeze replay even if freespace is zero

  5. Pass down information on table modification to scan nodes

  6. Thread flags through begin-scan APIs

  7. Remove PlannedStmt->resultRelations in favor of resultRelationRelids

  8. Make it cheap to check if a relation is modified by a query

  9. Keep newest live XID up-to-date even if page not all-visible

  10. Add pruning fast path for all-visible and all-frozen pages

  11. Detect and fix visibility map corruption in more cases

  12. Save vmbuffer in heap-specific scan descriptors for on-access pruning

  13. Avoid BufferGetPage() calls in heap_update()

  14. Initialize missing fields in CreateExecutorState()

  15. Prefix PruneState->all_{visible,frozen} with set_

  16. Add PageGetPruneXid() helper

  17. Move commonly used context into PruneState and simplify helpers

  18. Remove table_scan_analyze_next_tuple unneeded parameter OldestXmin

  19. Simplify visibility check in heap_page_would_be_all_visible()

  20. Rename LVRelState VM-related logging counters

  21. Fix flakiness in the pg_visibility VM-only vacuum test by using a temporary table.

  22. Refactor lazy_scan_prune() VM clear logic into helper

  23. Eliminate use of cached VM value in lazy_scan_prune()

  24. Combine visibilitymap_set() cases in lazy_scan_prune()

  25. Fix const qualification in prune_freeze_setup()

  26. Simplify vacuum visibility assertion

  27. Split heap_page_prune_and_freeze() into helpers

  28. Assert that cutoffs are provided if freezing will be attempted

  29. Split PruneFreezeParams initializers to one field per line

  30. Refactor heap_page_prune_and_freeze() parameters into a struct

  31. Make heap_page_is_all_visible independent of LVRelState

  32. Inline TransactionIdFollows/Precedes[OrEquals]()

  33. Add helper for freeze determination to heap_page_prune_and_freeze

  34. Bump XLOG_PAGE_MAGIC after xl_heap_prune change

  35. Correct prune WAL record opcode name in comment

  36. Add error codes when vacuum discovers VM corruption

  37. Remove unused xl_heap_prune member, reason

  38. Remove unneeded VM pin from VM replay

  39. Add assert and log message to visibilitymap_set

  40. Stamp HEAD as 19devel.

  41. Add error codes to some corruption log messages