Re: Table-level log_autovacuum_min_duration

On Fri, Mar 6, 2015 at 1:07 PM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Fri, Mar 6, 2015 at 12:44 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
>> On Thu, Mar 5, 2015 at 9:49 PM, Michael Paquier
>> <michael(dot)paquier(at)gmail(dot)com> wrote:
>>> On Thu, Mar 5, 2015 at 7:10 PM, Fujii Masao wrote:
>>>> With the patch, VACUUM ANALYZE VERBOSE doesn't emit any verbose message.
>>>> Why did you remove that functionality?
>>>
>>> Oops. Sorry about that. In gram.y, the combination of VacuumStmt with
>>> AnalyzeStmt overwrote the value of log_min_duration incorrectly. I
>>> also found another bug related to logging of ANALYZE not working
>>> correctly because of the use of IsAutoVacuumWorkerProcess() instead of
>>> VACOPT_VERBOSE (this is reducing the diffs of the patch btw). All
>>> those things are fixed in the attached.
>>
>> Thanks for updating the patch!
>>
>> Why does log_min_duration need to be set even when manual VACUUM command is
>> executed? Per the latest version of the patch, log_min_duration is checked only
>> when the process is autovacuum worker. So ISTM that log_min_duration doesn't
>> need to be set in gram.y. It's even confusing to me. Or if you're going to
>> implement something like "VACUUM VERBOSE DURATION n" (i.e., verbose message
>> is output if n seconds have been elapsed), that might be necessary, though...
>
> Thanks for reminding. The DURATION-like clause was exactly a point
> mentioned by Anzai-san upthread, and it made sense to me to be in-line
> with the other parameters controlling the freeze (discussion somewhat
> related to that =>
> http://www.postgresql.org/message-id/CAB7nPqRZX7Pv2B-R7xHmAh52tfjAQGfy9btqwFstgQgXks=iSw@mail.gmail.com)
> but we can live without it for this patch as VACOPT_VERBOSE is used
> only by manual VACUUM and not by autovacuum to choose the log elevel.

Are you planning to update the patch so that it's based on the commit 0d83138?

Regards,

--
Fujii Masao