Hi,
On 2024-04-12 09:43:46 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > Oh my. There's a workload that completely trivially hits this, without even
> > trying hard. LISTEN/NOTIFY.
>
> Hm. Bug in the NOTIFY logic perhaps?
I don't think it's a bug, async.c:SignalBackends() will signal once for every
NOTIFY, regardless of whether the target has already been signaled. You're
certainly right that:
> Sending that many signals can't be good from a performance POV, whether or
> not it triggers spinlock issues.
I wonder if we could switch this to latches, because with latches we'd only
re-send a signal if the receiving end has already processed the signal. Or
alternatively, make procsignal.c do something similar, although that might
take a bit of work to be done race-free.
Greetings,
Andres Freund