Re: Out-of-Cycle release? (was Re: BUG #19490: Streaming standby on 16.14 stops applying WAL on MultiXactOffsetSLRU when primary is 16.8)

Поиск
Список
Период
Сортировка
Искать

Re: Out-of-Cycle release? (was Re: BUG #19490: Streaming standby on 16.14 stops applying WAL on MultiXactOffsetSLRU when primary is 16.8)

От:
Andrey Borodin <x4mmm@yandex-team.ru>
Дата:


> On 25 Jun 2026, at 14:30, Michael Banck  wrote:
> 
> If either the standby gets updated to the
> latest minor release first, or if somebody does PITR from a primary on an
> earlier release to an instance with the latest minor release, the
> standby/PITR deadlocks and just sits there.

A bit of clarification. To trigger this bug WAL must be generated by 16.10
or prior and replay must be done by latest release (16.14). And the WAL
trace must contain sufficient number of MultiXacts.

So far we had 3 or 4 reports on pgsql-bugs and few in chats.

But yeah, restoration from an old backup is a typical use case.


Best regards, Andrey Borodin.


Re: Out-of-Cycle release? (was Re: BUG #19490: Streaming standby on 16.14 stops applying WAL on MultiXactOffsetSLRU when primary is 16.8)

От:
Andrey Borodin <x4mmm@yandex-team.ru>
Дата:


> On 25 Jun 2026, at 15:14, Andrey Borodin  wrote:
> 
> WAL must be generated by 16.10
> or prior

Sorry for the noise, but WAL from 16.11 also can trigger this problem.


Best regards, Andrey Borodin.


pgsql: Fix GIN posting tree page deletion with incomplete splits.

От:
Peter Geoghegan <pg@bowt.ie>
Дата:
Fix GIN posting tree page deletion with incomplete splits.

GIN posting tree page deletion failed to consider whether the target
page's left sibling page, or the deletion target itself, was marked as
incompletely split.  Page deletion finds the target page's left sibling
by walking the parent's downlinks, but an incompletely split page's new
right half is part of the sibling chain despite having no downlink.
Deletion could therefore overwrite the rightlink of the wrong page,
disconnecting the split's still-live right half from the sibling chain.
Scans would then silently miss tuples from that page.

To fix, teach the relevant page deletion path to avoid deleting a
posting tree page whose left sibling is marked incompletely split (and
to avoid doing so when the target page itself is so marked).  This is
essentially the same approach used by nbtree page deletion.

Claude Code found this problem.  The committed test case is a simplified
version of the one that it wrote to demonstrate this bug.

Author: Peter Geoghegan 
Reviewed-by: Andrey Borodin 
Discussion: https://postgr.es/m/CAH2-Wz=sKJcn+OtfVN9rdg+Ps9e4cuQWNP-9t12UE2d8nEG90Q@mail.gmail.com
Backpatch-through: 14

Branch
------
REL_14_STABLE

Details
-------
https://git.postgresql.org/pg/commitdiff/bbe7aca31db0b9e4758bb67a2d42f1bcc05e0c4b

Modified Files
--------------
src/backend/access/gin/ginvacuum.c | 19 +++++++++++++++++--
1 file changed, 17 insertions(+), 2 deletions(-)

FAQ