Re: Does this perf output seem 'normal'?

On 27.04.2016 2:27, Peter Devoy wrote:
> Hi all,
>
> I am trying to work out why a piece of software, Mapnik, is executing
> slowly. All it is doing is loading a config file which causes about
> 12 preparation queries (i.e. with LIMIT 0) to be executed. I can see
> from pg_stat_statements these only take ~1ms in their totality.
>
> So next I ran "perf record -a -g sleep 0.4" and quickly loaded my PHP
> page to start the process (I am interfacing with mapnik via a PHP
> Extension), this is a sample of the result:
>
> https://cloud.githubusercontent.com/assets/4375326/14836015/08b0a8b4-0c05-11e6-96e5-f1836497cdfe.png
>
> The PHP page returns in ~60ms (well within the 400ms used for the perf
> recording). If possible, please can anyone with more familiarity with
> perf and Postgres tell me whether the result of the recording looks
> healthy?
>
> I am wondering if there is some problem with the way in which Mapnik
> is implementing libpq -- I already moved from Mapnik 2.2 because it
> was effectively DOSing Postgres by leaving client connections open.
>
> Any help would be greatly appreciated.
>
>
> Kind regards
>
> Peter
>
>
Hi! What do you want to see in perf stats? Maybe you can explain your
problem more in details?

--
Alex Ignatov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company