[Labs-l] Performance at tool-labs compared with TS

[Petr Bena]

I wanted to know which resources your tool requires most (if cpu or
mysql or some remote api or filesystem... etc) that can help us track
what is causing this slowness. There are things labs are surely faster
in than toolserver, as well as things they are probably slower in. We
need to find out which parts needs the improvement.

On Sun, May 5, 2013 at 4:26 PM, Kolossos <tim.alder at s2002.tu-chemnitz.de> wrote:
> I'm not sure for what it should be useful to spend hours to move everything
> to bots, if the toolserver is a running reference. The tool is running since
> 2007 so it's not anymore beta.
> Btw: I don't understand the documentation of "bots", somebody should clean
> it up, so that I know where to start.
>
> My idea is more to simplify the sql query to make a new benchmark, but today
> the toolserver is not stable enough for this, I get a lot of 502.
>
> The answer for your question about db, filesystem, cpu, memory, remote api
> is here:
> https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools
> and
> https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/Help
> or not?
> I simply following the documentation.
>
> Greetings Kolossos
>
>
> Am 05.05.2013 11:08, schrieb Petr Bena:
>
>> Hi, could you do similar test on bots project? That is supposed to be
>> a staging area - so in future it will be tools beta version.
>>
>> There are very much lower restrictions than on tools project, so we
>> might see if this problem is related to performance of VM or if that
>> is related to configuration. You don't need to even create a tool
>> account there, you just ssh, upload your tool and that is all. There
>> are no resource limits at all
>>
>> On Sun, May 5, 2013 at 1:37 AM, Kolossos <tim.alder at s2002.tu-chemnitz.de>
>> wrote:
>>>
>>> Hello,
>>> I move my templatetiger scripts as first tool from toolserver to labs. I
>>> compared performance of toolserver script[1] with that on labs[2]. To my
>>> suprise the labs script needs significant longer. So toolserver script
>>> needs
>>> 0.5 sec and labs script needs 5 sec. (Factor 10 !) If I increase the
>>> limit
>>> from 30 to 300 the toolserver script needs 2 seconds, on labs it need
>>> unacceptable 48 sec.
>>>
>>> So something is going wrong.
>>>
>>> I test a php script with a simple but long for-next-loop, both servers
>>> react
>>> similar, so php seem not be the problem.
>>> The queries directly on mysql-terminal are fast and seems not to be the
>>> problem. So perhaps the interface is the problem. Other reason could be
>>> that
>>> I work with SQL_CALC_FOUND_ROWS because Mysql had 2007 a bug with
>>> sub-queries, perhaps this makes now problems at maria.
>>>
>>> The php-script is simply a copy and I import the extracts of Alemannic to
>>> the database in the same way. To load at toollabs were with 0.2 relative
>>> low.
>>>
>>> What's the reason? Could somebody verify the behaviour I found.
>>>
>>> Greetings Tim alias Kolossos
>>>
>>>
>>> [1]http://toolserver.org/~kolossos/templatetiger/tt-table4.php?lang=alswiki&template=Gemeinde%20in%20Frankreich&where=&is=&&offset=240&limit=30
>>>
>>>
>>> [2]http://tools.wmflabs.org/templatetiger/tt-table4.php?lang=alswiki&template=Gemeinde%20in%20Frankreich&where=&is=&&offset=30&limit=30
>>>
>>>
>>> _______________________________________________
>>> Labs-l mailing list
>>> Labs-l at lists.wikimedia.org
>>> https://lists.wikimedia.org/mailman/listinfo/labs-l
>>
>>
>> _______________________________________________
>> Labs-l mailing list
>> Labs-l at lists.wikimedia.org
>> https://lists.wikimedia.org/mailman/listinfo/labs-l
>>
>
>
>
> _______________________________________________
> Labs-l mailing list
> Labs-l at lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/labs-l