jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Antonio Gomes Rodrigues <ra0...@gmail.com>
Subject Re: Report / Dashboard : New report containing min/max and percentiles graphs / BUG 60091
Date Wed, 07 Sep 2016 12:25:05 GMT
My opinon and use (for report and SLA but not for troubleshooting) of TC is
to hide requests and work with "functional" requests

For example if my login process take 3 requests, in my report I want to
show the response time of the response time TC and not of the 3 requests

Same if I want to compare 2 releases (e.g in release 1 I have 3 requests to
make the login and in release 2, I have only 2 requests. In this example
it's easiest to compare TC)

To troubleshooting, I need to have response time/code return... of each
request

Antonio

2016-09-07 14:15 GMT+02:00 Philippe Mouawad <philippe.mouawad@gmail.com>:

> On Wednesday, September 7, 2016, Antonio Gomes Rodrigues <ra0077@gmail.com
> >
> wrote:
>
> > Hi,
> >
> > See my answer below
> >
> > 2016-09-06 21:28 GMT+02:00 Philippe Mouawad <philippe.mouawad@gmail.com
> > <javascript:;>>:
> >
> > > Hello,
> > > I commited few minutes ago the implementation of a new Graph in the
> HTML
> > > report:
> > > - https://bz.apache.org/bugzilla/show_bug.cgi?id=60091
> > >
> > > As always feedback is welcome.
> > > But I have 2 questions:
> > > 1/ on how percentile should be computed.
> > > For now percentile is computed per each time period as defined by
> > > granularity, starting from nothing for each time period.
> > > Should we instead take the last percentile computed for previous time
> > > period or proceed differently ?
> > >
> >
> > No opinion
>
>
>
>
> >
> >
> > >
> > > 2/ Should we exclude Transaction Controller from computation ? Because
> > they
> > > tend if included to hide real response times as being a group of
> > requests,
> > > there value will be much higher than response times.
> > >
> >
> > In y opinion, we should not mix TC and requests grouped by TC but take
> > account of TC usage
> >
> > example 1:
> > TC1:
> >        request 1
> >        request 2
> > TC2:
> >        request 3
> >        request 4
> >
> > Here what I want to see is a Graph that contains min, max, percentil...
> for
> > only TC (TC1 + TC2)
>
>
> you would check generate parent sample here, right ?
>
> > Or if it not possible a Graph that contains min, max, percentil... for
> only
> > requests (request 1 + request 2 + request 3 + request 4)
>
>
> Excluding tc is the easiest way.
>
> The problem we have is that report requires tc to not generate a parent
> sample otherwise we loose important information:
> - error codes of underlying requests
> - number of request to compute hits
> - precise response time for each request
>
> But if we don't generate parent sample:
> - errors can count twice(for the child request and its parent TC)
> - percentiles are also impacted as you describe it
>
> TC are in fact a pain in the ass in Jmeter code.
> But they are still useful to group requests.
>
>
>
>
>
>
>
> > example 2:
> > TC1:
> >        request 1
> >        request 2
> > request 3
> > request 4
> >
> > Here what I want to see is a Graph that contains min, max, percentil...
> for
> > TC1 + request 3 + request 4
>
>
> So you would check generate parent sample here ?
>
> >
> >
> > Antonio
> >
> >
> > > Thank you
> > >
> > > --
> > > Regards.
> > > Philippe M.
> > >
> >
>
>
> --
> Cordialement.
> Philippe Mouawad.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message