Re: Table-level log_autovacuum_min_duration

On Mon, Mar 23, 2015 at 1:54 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
> On Thu, Mar 19, 2015 at 1:43 PM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> On Thu, Mar 19, 2015 at 12:40 PM, Michael Paquier
>> <michael(dot)paquier(at)gmail(dot)com> wrote:
>>> On Thu, Mar 19, 2015 at 12:23 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
>>>> Are you planning to update the patch so that it's based on the commit 0d83138?
>>>
>>> Yes... Very soon.
>>
>> And here is the rebased patch.
>
> Thanks for rebasing the patch! Looks good to me.
>
> One concern about this patch is; currently log_autovacuum_min_duration can be
> changed even while autovacuum worker is running. So, for example, when
> the admin notices that autovacuum is taking very long time, he or she can
> enable logging of autovacuum activity on the fly. But this patch completely
> prevents us from doing that, because, with the patch, autovacuum worker always
> picks up the latest setting value at its starting time and then keeps using it
> to the end. Probably I can live with this. But does anyone has other thought?

In AutoVacWorkerMain, I am reading the following:

* Currently, we don't pay attention to postgresql.conf changes that
* happen during a single daemon iteration, so we can ignore SIGHUP.
*/
pqsignal(SIGHUP, SIG_IGN);

So a worker does not see changes in postgresql.conf once it is run and
processes a database, no? The launcher does run ProcessConfigFile()
when SIGHUP shows up though.
Regards,
--
Michael