Stress test to trytond

[Luciano Rossi]

Hi everybody,

I have my own server on the cloud. I want to do some stress tests to my
trytond. I did not found any information about that.

I want to generate some metrics, so if I changed any configuration of
the trytond or postgresql I can see if there was any differences.

any idea?

Thanks!

--
Luciano Rossi - lu...@gcoop.coop
http://gcoop.coop - Cooperativa de Software Libre
Julián Alvarez 958 CABA C1414DRT +54 11 4771-0055

Visitá www.tiendas.coop - www.nube.coop

[Cédric Krier]

On 2017-06-30 13:03, Luciano Rossi wrote:
> Hi everybody,
>
> I have my own server on the cloud. I want to do some stress tests to my
> trytond. I did not found any information about that.
>
> I want to generate some metrics, so if I changed any configuration of
> the trytond or postgresql I can see if there was any differences.

I guess you will have to decide exactly what you want to test and write
a scenario for it.

I think the more accurate will be a scenario that does the most common
things (like: create sale, ship product, validate invoice).
You can write such test with proteus and run this scenario multiple
times in parallel started at different times over a period and collect
the time to complete for each one. Then you can make a mean/standard
deviation etc.

Another kind of test would be to pick an simple operation like reading
the fields of a record. And use a stress tools to repeat this operation
as much as possible.

A last kind of test would be about large data. You could forge a request
that require a lot of resources like read a lot of records or a search
that return a lot of records etc.

Of course, those tests will depend not only of the trytond core but also
of course of the database and the eventual proxy server. But also it
will depend of the implementation of the model used. For example, poor
implemented Function getter will kill the performance.
Also you should be careful that trytond and the database may have some
warm-up time because of the cache, so the measures should take care of
that.

If you could create any of those tests in reliable way, it will be good
to share and to compare different version of Tryton to see if we have
regression, where are the bottlenecks and what are the best setups per
cases.

--
Cédric Krier - B2CK SPRL
Email/Jabber: cedric...@b2ck.com
Tel: +32 472 54 46 59
Website: http://www.b2ck.com/

[Cédric Krier]

On 2017-07-03 01:07, Cédric Krier wrote:
> On 2017-06-30 13:03, Luciano Rossi wrote:
> > Hi everybody,
> >
> > I have my own server on the cloud. I want to do some stress tests to my
> > trytond. I did not found any information about that.
> >
> > I want to generate some metrics, so if I changed any configuration of
> > the trytond or postgresql I can see if there was any differences.
>
> I guess you will have to decide exactly what you want to test and write
> a scenario for it.
>
> I think the more accurate will be a scenario that does the most common
> things (like: create sale, ship product, validate invoice).
> You can write such test with proteus and run this scenario multiple
> times in parallel started at different times over a period and collect
> the time to complete for each one. Then you can make a mean/standard
> deviation etc.
>
> Another kind of test would be to pick an simple operation like reading
> the fields of a record. And use a stress tools to repeat this operation
> as much as possible.
>
> A last kind of test would be about large data. You could forge a request
> that require a lot of resources like read a lot of records or a search
> that return a lot of records etc.

I forgot to say that proteus is not really designed to work efficiently
with large data. It fetch record values record by record instead of
using a call with a list of ids.

> Of course, those tests will depend not only of the trytond core but also
> of course of the database and the eventual proxy server. But also it
> will depend of the implementation of the model used. For example, poor
> implemented Function getter will kill the performance.
> Also you should be careful that trytond and the database may have some
> warm-up time because of the cache, so the measures should take care of
> that.
>
> If you could create any of those tests in reliable way, it will be good
> to share and to compare different version of Tryton to see if we have
> regression, where are the bottlenecks and what are the best setups per
> cases.

It will be also good to test with different version of Python.


The tests machine could be used to run such benchmark tool.