Horton
Menu

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

Andrey Borodin <x4mmm@yandex-team.ru>

View thread
From: Andrey Borodin <x4mmm@yandex-team.ru>
To: Melanie Plageman <melanieplageman@gmail.com>
Cc: Kirill Reshke <reshkekirill@gmail.com>, Chao Li <li.evan.chao@gmail.com>, Xuneng Zhou <xunengzhou@gmail.com>, Andres Freund <andres@anarazel.de>, Robert Haas <robertmhaas@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Heikki Linnakangas <hlinnaka@iki.fi>
Date: 2026-01-16T13:05:44Z
Lists: pgsql-hackers 

> 
>> In a step "Pass down information on table modification to scan node" you pass SO_HINT_REL_READ_ONLY flag in IndexNext() and BitmapTableScanSetup(), but not in IndexNextWithReorder() and IndexOnlyNext(). Is there a reason why index scans with ordering cannot use on-access VM setting?
> 
> Great point, I simply hadn't tested those cases and didn't think to
> add them. I've added them in attached v33.
> 
> While looking at other callers of index_beginscan(), I was wondering
> if systable_beginscan() and systable_beginscan_ordered() should ever
> pass SO_HINT_REL_READ_ONLY. I guess we would need to pass if the
> operation is read-only above the index_beginscan() -- I'm not sure if
> we always know in the caller of systable_beginscan() whether this
> operation will modify the catalog. That seems like it could be a
> separate project, though, so maybe it is better to say this feature is
> just for regular tables.

+1 for excluding catalog. 

> 
> As for the other cases: We don't have the relation range table index
> in check_exclusion_or_unique_constraints(), so I don't think we can do
> it there.
> 
> And I think that the other index scan cases like in replication code
> or get_actual_variable_endpoint() are too small to be worth it, don't
> have the needed info, or don't do on-access pruning (bc of the
> snapshot type they use).

Makes sense. I think with cases that you already added the list is complete.


> On 6 Jan 2026, at 22:31, Melanie Plageman <melanieplageman@gmail.com> wrote:
> 
> Let me know if there
> are other patches you reviewed that you did not mention.

I'll review them one by one, starting from first 3.

> <v33-0001-Combine-visibilitymap_set-cases-in-lazy_scan_pru.patch>

I agree that cases of visibilitymap_set() are merged properly.

I'm slightly worried that now if we decided to visibilitymap_set() all_frozen page it goes before corruption warning in block of

else if (all_visible_according_to_vm && !PageIsAllVisible(page) &&
			 visibilitymap_get_status(vacrel->rel, blkno, &vmbuffer) != 0)
{
}

Perhaps, it's not that important though. We are resetting everything to correct state anyway, just without WARNING.

> <v33-0002-Eliminate-use-of-cached-VM-value-in-lazy_scan_pr.patch>

LGTM. I'm actually happy with "else if (!PageIsAllVisible(page) &&..)" conversion to "if (!PageIsAllVisible(page) &&..)".

> <v33-0003-Refactor-lazy_scan_prune-VM-clear-logic-into-hel.patch>


LGTM.

I'll proceed with other steps later. Thanks!


Best regards, Andrey Borodin.

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