Horton
Menu

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

Alexey Makhmutov <a.makhmutov@postgrespro.ru>

View thread
From: Alexey Makhmutov <a.makhmutov@postgrespro.ru>
To: Andres Freund <andres@anarazel.de>
Cc: Melanie Plageman <melanieplageman@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Kirill Reshke <reshkekirill@gmail.com>, Andrey Borodin <x4mmm@yandex-team.ru>, Robert Haas <robertmhaas@gmail.com>, Heikki Linnakangas <hlinnaka@iki.fi>
Date: 2026-04-06T15:32:03Z
Lists: pgsql-hackers 
Hi Andres,

On 4/6/26 17:44, Andres Freund wrote:
> I don't have a strong opinion on this, but I think it's pretty defensible to
> record only when there's free space. The whole goal of updating the FSM during
> recovery is to make sure that free space can be found fairly quickly after
> promotion (it's also beneficial in some crash recovery cases, but not that
> much).
> ... 
> Obviously the FSM is not crashsafe, so updating it with 0 during replay could
> avoid some unnecessary page reads after a promotion. But I'm not sure that
> that's particularly worth optimizing for.

This makes sense. I'd like just to put some context here: I was checking 
the FSM update case in scope of the thread 
https://www.postgresql.org/message-id/flat/596c4f1c-f966-4512-b9c9-dd8fbcaf0928@postgrespro.ru, 
in which I was specifically looking at the case with outdated FSM data 
(showing lots of free space) on standby causing a significant 
performance hit after switchover. As example this include case with 
table having fillfactor<=80 which has prior bulk rows deletes + 
insertion. In this case (mostly) empty FSM block may be delivered to 
standby via FPI, but subsequent inserts may be lost due to the 20% 
heuristic. Moreover, updates to FSM may be lost even for blocks filled 
for more than 80% due to missing dirty flag as described in that thread.

In my understanding the FSM update during processing of the 
'heap_xlog_visible' function on standby was kind of 'last line of 
defense' for any corner case scenario with FSM update (as block would 
not be visited by the vacuum process once it's marked as 'all visible') 
and it was introduced in 
https://www.postgresql.org/message-id/20180802172857.5skoexsilnjvgruk@alvherre.pgsql 
(ab7dbd681) specifically for this purpose. Now, as logic of 
'heap_xlog_visible' is merged into 'heap_xlog_prune_freeze', so this 
task is carried by this function.

I fully agree that having exactly zero-space seems to be a very uncommon 
situation (and probably not reproducible with tables having 
fillfactor<=80). I've just noted that such case was processed by the old 
logic in the 'heap_xlog_visible', while current implementation in 
'heap_xlog_prune_freeze' skips it.

Thanks,
Alexey

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