"Save idle time" by default

[Vladimir Sitnikov <si...@gmail.com>]

Hi,

I analyze jmeter csv raw results via both jmeter listeners (e.g.
AggregateReport) and build my own reports via r.

It works pretty much fine, however I faced an issue with transaction
controllers and timers.

Here's the issue: by default the log (e.g. raw log in csv) does _not_
include duration of the timers, so there is no way to tell when
transaction finished.
There is a similar problem with regular samplers: it is hard to tell
why there is a gap between two samplers.

One of my reports is a sequence chart of jmeter transactions with
actions in the system under test, thus I need precise information on
the start and end timestamps.
Gaps between JMeter requests bothers me since I do not know if they
are caused by JMeter or by a bad script or whatever else.

I've scanned through jmeter.properties and identified that
jmeter.save.saveservice.idle_time=true solves solves the problem to a
certain degree.
With this option, log for transaction controller does report delay time.
However, there still no way to map those delays to the timeline:
individual samplers always show 0 in the "idle time column"

Can you please clarify:
1) Why "save idle time" is not enabled by default?
2) Is there an option to enable logging of "preprocessors duration"
and "postprocessors duration" for the samplers?
If JMeter could log those timings it might have a good purpose:
identify inefficient regexps, eliminate surprises when analyzing "gaps
in between requests", etc.

-- 
Regards,
Vladimir Sitnikov

Re: "Save idle time" by default

[Vladimir Sitnikov <si...@gmail.com>]

>Do you think it is useful for you ?

I do.

As far as I see, enabling idle_time=true does not add new
System.currentTimeMillis()
kind of calls, thus it should be relatively safe performance-wise.

>More work, patch are welcome for 3.1 release maybe, I will be happy to
>integrate it.

It makes sense.

Vladimir

Re: "Save idle time" by default

[Philippe Mouawad <ph...@gmail.com>]

Hi Vladimir,
Answers below.

On Sat, Feb 27, 2016 at 10:53 AM, Vladimir Sitnikov <
sitnikov.vladimir@gmail.com> wrote:

> Do you mean just idle_time=true flip?
>
Yes for 3.0.0.
Do you think it is useful for you ?
if not we hold on this one to include below notes.

Do you consider to include "delay time" to the logged "idle_time" of
> simple samplers?
>

More work, patch are welcome for 3.1 release maybe, I will be happy to
integrate it.

> Do you consider adding preprocessors duration column?
>
No in the scope of this one as per previous note.

>
> Vladimir
>



-- 
Cordialement.
Philippe Mouawad.

Re: "Save idle time" by default

[Vladimir Sitnikov <si...@gmail.com>]

Do you mean just idle_time=true flip?
Do you consider to include "delay time" to the logged "idle_time" of
simple samplers?
Do you consider adding preprocessors duration column?

Vladimir

Re: "Save idle time" by default

[Philippe Mouawad <ph...@gmail.com>]

Hi ,
Shall I change this for 3.0 ?
Thanks

On Fri, Oct 31, 2014 at 6:16 PM, Philippe Mouawad <
philippe.mouawad@gmail.com> wrote:

>
>
> On Friday, October 31, 2014, Vladimir Sitnikov <
> sitnikov.vladimir@gmail.com> wrote:
>
>> >You can do so as you explained no ?
>> I can.
>> I just do not agree with "But regarding this one I would say that
>> under load test ( once you're sure
>> you don't have issues in your plan you are not interested by this info)."
>> I am going to have this option enabled, especially for load tests.
>> Without load JMeter usually is perfect in terms of "no gaps, etc".
>>
>> Agree now :) will update this if nobody objects
>
>> >You mean you want this enabled by default ?
>> Yes. I guess it will not impact me much since I've included this
>> setting to the list of "must be configured before any JMeter usage",
>> however I feel this might improve the product in general.
>>
>> > Yes, it could be
>> Ok, will see if I can figure out the patch.
>>
>> Thanks
>
>> > important known bug 56160
>> Thank you. Will check that.
>>
>> -
>> Vladimir
>>
>
>
> --
> Cordialement.
> Philippe Mouawad.
>
>
>
>


-- 
Cordialement.
Philippe Mouawad.

Re: "Save idle time" by default

[Philippe Mouawad <ph...@gmail.com>]

On Friday, October 31, 2014, Vladimir Sitnikov <si...@gmail.com>
wrote:

> >You can do so as you explained no ?
> I can.
> I just do not agree with "But regarding this one I would say that
> under load test ( once you're sure
> you don't have issues in your plan you are not interested by this info)."
> I am going to have this option enabled, especially for load tests.
> Without load JMeter usually is perfect in terms of "no gaps, etc".
>
> Agree now :) will update this if nobody objects

> >You mean you want this enabled by default ?
> Yes. I guess it will not impact me much since I've included this
> setting to the list of "must be configured before any JMeter usage",
> however I feel this might improve the product in general.
>
> > Yes, it could be
> Ok, will see if I can figure out the patch.
>
> Thanks

> > important known bug 56160
> Thank you. Will check that.
>
> -
> Vladimir
>


-- 
Cordialement.
Philippe Mouawad.

Re: "Save idle time" by default

[Vladimir Sitnikov <si...@gmail.com>]

>You can do so as you explained no ?
I can.
I just do not agree with "But regarding this one I would say that
under load test ( once you're sure
you don't have issues in your plan you are not interested by this info)."
I am going to have this option enabled, especially for load tests.
Without load JMeter usually is perfect in terms of "no gaps, etc".

>You mean you want this enabled by default ?
Yes. I guess it will not impact me much since I've included this
setting to the list of "must be configured before any JMeter usage",
however I feel this might improve the product in general.

> Yes, it could be
Ok, will see if I can figure out the patch.

> important known bug 56160
Thank you. Will check that.

-
Vladimir

Re: "Save idle time" by default

[Philippe Mouawad <ph...@gmail.com>]

On Fri, Oct 31, 2014 at 5:20 PM, Vladimir Sitnikov <
sitnikov.vladimir@gmail.com> wrote:

> >But regarding this one I would say that under load test
>
> I want to capture enough information to be able to analyze gaps if
> load test results in some gaps.
> It looks like this information does not involve high overhead, thus I
> would like to have it enabled just in case.
>

You can do so as you explained no ?
You mean you want this enabled by default ?

>
> >>2) Is there an option to enable logging of "preprocessors duration"
> >> and "postprocessors duration" for the samplers?
> >No, it would require a patch.
>
> Do you feel this patch might be accepted in the main codebase?
>
> Yes, it could be a way to detect Coordinate Omission problems.

> -
> Vladimir
>



-- 
Cordialement.
Philippe Mouawad.

Re: "Save idle time" by default

[Vladimir Sitnikov <si...@gmail.com>]

>But regarding this one I would say that under load test

I want to capture enough information to be able to analyze gaps if
load test results in some gaps.
It looks like this information does not involve high overhead, thus I
would like to have it enabled just in case.

>>2) Is there an option to enable logging of "preprocessors duration"
>> and "postprocessors duration" for the samplers?
>No, it would require a patch.

Do you feel this patch might be accepted in the main codebase?

-
Vladimir

Re: "Save idle time" by default

[Philippe Mouawad <ph...@gmail.com>]

On Friday, October 31, 2014, Vladimir Sitnikov <si...@gmail.com>
wrote:

> Hi,
>
> I analyze jmeter csv raw results via both jmeter listeners (e.g.
> AggregateReport) and build my own reports via r.
>
> It works pretty much fine, however I faced an issue with transaction
> controllers and timers.
>
> Here's the issue: by default the log (e.g. raw log in csv) does _not_
> include duration of the timers, so there is no way to tell when
> transaction finished.
> There is a similar problem with regular samplers: it is hard to tell
> why there is a gap between two samplers.
>
> One of my reports is a sequence chart of jmeter transactions with
> actions in the system under test, thus I need precise information on
> the start and end timestamps.
> Gaps between JMeter requests bothers me since I do not know if they
> are caused by JMeter or by a bad script or whatever else.
>
> I've scanned through jmeter.properties and identified that
> jmeter.save.saveservice.idle_time=true solves solves the problem to a
> certain degree.
> With this option, log for transaction controller does report delay time.
> However, there still no way to map those delays to the timeline:
> individual samplers always show 0 in the "idle time column"
>
> Can you please clarify:
> 1) Why "save idle time" is not enabled by default?


Sometimes there is no real explanation, so in recent versions we tried and
continue to improve the default settings (for example
jmeter.save.saveservice.thread_counts will be saved starting from 2.12) but
task is still in progress.
But regarding this one I would say that under load test ( once you're sure
you don't have issues in your plan you are not interested by this info).
So as long as you can change it in user.properties it seems fine to me.
Maybe needs to be better documented if it took you some time to find it.

2) Is there an option to enable logging of "preprocessors duration"
> and "postprocessors duration" for the samplers?


No, it would require a patch.


If JMeter could log those timings it might have a good purpose:
> identify inefficient regexps, eliminate surprises when analyzing "gaps
> in between requests", etc.
>
> Note that in future 2.12, an important known bug 56160 has been fixed that
could create gaps between requests if it was met.
Try nightly build if your gaps seem abnormal.


> --
> Regards,
> Vladimir Sitnikov
>