Re: PostgreSQL 18.6/17.11: standby PANIC on restart after VM truncation

From: Melanie Plageman <melanieplageman(at)gmail(dot)com>
To: shihao zhong <zhong950419(at)gmail(dot)com>
Cc: Rahul <rahul(at)rhyadav(dot)dev>, Kirill Reshke <reshkekirill(at)gmail(dot)com>, Nktpro <nktpro(at)gmail(dot)com>, Pgsql Bugs <pgsql-bugs(at)lists(dot)postgresql(dot)org>
Subject: Re: PostgreSQL 18.6/17.11: standby PANIC on restart after VM truncation
Date: 2026-11-16 16:03:25
Message-ID: CAAKRu_bnmOWQiQN93qS1oEEVw8X09Y-sGhs56avw8y7ysffw1w@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

On Tue, Oct 13, 2026 at 11:20 PM shihao zhong <zhong950419(at)gmail(dot)com> wrote:
>
> > Jacky's original fix with XLogTruncateRelation() is the one I prefer.
>
> Fine with me. I tested v5 on master, 19, 18 and 17. With the fix the
> standby restarts.

While working on committing this, an AI review pointed out that the
HEAPBLK_TO_MAPBLOCK_LIMIT() macro is not overflow safe and needs to be
fixed. This patch adds a user of visibilitymap_truncation_length(),
but the existing user, wal summarizer can already encounter this bug.
A heap block number near MaxBlockNumber can be passed it and then
adding HEAPBLOCKS_PER_PAGE - 1 overflows it and wraps it to a small
number, recording a VM limit much smaller than is correct and causing
the summarizer to discard tracked VM modifications for the rest of the
fork.

I think the attached preliminary patch 0001 has to be committed before 0002.

We don't have a good way of adding a regression test for this that
doesn't take forever because of the large heap block number required.

> One small thing. On 17 the test part does not apply, because
> 001_concurrent_transaction.pl has no standby there.

Thanks for the heads up. I've implemented the 17 port locally.

- Melanie

Attachment Content-Type Size
v6-0001-Avoid-overflow-computing-visibility-map-length.patch text/x-patch 1.5 KB
v6-0002-Forget-invalid-VM-pages-during-truncation-replay.patch text/x-patch 4.8 KB

In response to

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message Matheus Alcantara 2026-11-16 16:07:34 Re: BUG #19772: ERROR: corrupt MVNDistinct entry for GROUP BY c0, ROLLUP (abs(c0)) with ndistinct statistics
Previous Message Dilip Kumar 2026-11-16 15:46:23 Re: BUG #19892: Logical replication skips UPDATE/DELETE of rows with NULL in a hash-indexed subscriber column