Hi,
On Wed, Jan 28, 2026 at 03:35:21PM +0900, Michael Paquier wrote:
> On Mon, Jan 26, 2026 at 06:59:28AM +0000, Bertrand Drouvot wrote:
> > The attached, to apply on top of 0001, fix the issue. However it handles only the
> > WaitLatch in ProcSleep() case and I start to have concern about the others WaitLatch()
> > that would/could be "woken up" every 1s.
>
> > Using disable_timeout() and enable_timeout_after() in WaitEventSetWait() does not
> > look like a great answer to this concern, so I wonder if we should use a larger
> > flush frequency instead (as proposed up-thread), thoughts?
>
> Only a larger frequency is not the correct answer here. It would just
> reduce the frequency of the extra lock wait messages for one: these
> should never appear more than necessary.
Right. The fix in fix_ProcSleep.txt shared up-thread solves that and has been
added in 0001 attached.
Also the attached is now split in 4 sub-patches with 0002 introducing a new
GUC to control the flush interval (default is 10s). Note that 0001 to 0003 could
be merged as one patch but I did it that way to ease the review.
The new version also adds more documentation and takes care of Sami's comments
shared up-thread.
> And how about for example extension code?
I think that depending on how they write their code around WaitLatch (if any) they
could see messages being reported (if they do) in the logs at stats_flush_interval
frequency.
That said the default value is 10s and that looks pretty long for a latch to wait on.
In any case, they should already take care of a latch being woken by WL_LATCH_SET.
What do you think?
Regards,
--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
