Le 28/10/2019 à 17:25, Tom Lane a écrit :
> =?UTF-8?Q?Gr=c3=a9goire_de_Turckheim?= <gdeturckheim@scaleway.com> writes:
>> Le 28/10/2019 à 15:22, Tom Lane a écrit :
>>> We made some performance improvements for NOTIFY just a couple months
>>> ago, cf commits b10f40bf0, bb5ae8f6c, bca6e6435, 51004c717. It would
>>> be interesting to know how much those changes helped your use-case.
>> If my understanding of the problem is correct, there is no performance
>> issue with the notification itself.
>> The problem is the following: a system-wide lock is taken pre-commit, so
>> any other transaction with a NOTIFY will have to wait for other
>> transactions to complete before it can leave its own pre-commit stage.
> Right, but all commits are single-threaded at some granularity.
> The big problem with NOTIFY is that it sits for a long time holding
> that lock, if you have a lot of notify traffic. The commits I mentioned
> should improve that.
>
> Anyway, as I said, it would be good to find out whether the already
> finished fixes are enough to solve your problem, before we debate
> whether more needs to be done.
Let's do it this way, I'll give it a try (might be long, this isn't
something we can easily upgrade) and get back to you.
Thanks!
--
Grégoire de Turckheim