Hi, Alexander!
On Mon, Feb 5, 2024 at 7:00 PM Alexander Lakhin <exclusion@gmail.com> wrote:
> 05.02.2024 02:51, Alexander Korotkov wrote:
>
> > Usage of heap_inplace_update() seems appropriate for me here. It
> > avoids trouble with both TOAST and row-level locks. Alexander, could
> > you please recheck this fixes the problem.
>
> I've re-tested the last problematic scenario and can confirm that the fix
> works for it (though it still doesn't prevent the autovacuum issue (with
> 4b885d01 reverted)), but using heap_inplace_update() was considered risky
> in a recent discussion:
> https://www.postgresql.org/message-id/1596629.1698435146%40sss.pgh.pa.us
> So maybe it's worth to postpone such a fix till that discussion is
> finished or to look for another approach...
Thank you for pointing this out. I don't think there is a particular
problem with this use case for the following reasons.
1) Race conditions over pg_database.dathasloginevt are protected with lock tag.
2) Unsetting pg_database.dathasloginevt of old tuple version shouldn't
cause a problem. The next tuple version will be updated by further
connections.
However, I agree that it's better to wait for the discussion you've
pointed out before introducing another use case of
heap_inplace_update().
------
Regards,
Alexander Korotkov