Re: Suppressing useless wakeups in walreceiver

On Tue, Oct 11, 2022 at 07:01:26AM +0530, Bharath Rupireddy wrote:
> On Mon, Oct 10, 2022 at 11:21 PM Nathan Bossart
>> Outside of the code snippet you pointed out,
>> I think WalReceiverMain() has a similar problem. That being said, I'm
>> generally skeptical that this sort of thing is detrimental given the
>> current behavior (i.e., wake up every 100ms), the usual values of these
>> GUCs (i.e., tens of seconds), and the fact that any tasks that are
>> inappropriately skipped will typically be retried in the next iteration of
>> the loop.
>
> AFICS, the aim of the patch isn't optimizing around
> GetCurrentTimestamp() calls and the patch does seem to change quite a
> bit of them which may cause a change of the behaviour. I think that
> the GetCurrentTimestamp() optimizations can go to 0003 patch in this
> thread itself or we can discuss it as a separate thread to seek more
> thoughts.
>
> The 'now' in many instances in the patch may not actually represent
> the true current time and it includes time taken by other operations
> as well. I think this will have problems.

What problems do you think this will cause?

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com