jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From sebb <seb...@gmail.com>
Subject Re: Mailer Visualizer
Date Sun, 26 Feb 2017 12:44:21 GMT
On 26 February 2017 at 12:33, Philippe Mouawad
<philippe.mouawad@gmail.com> wrote:
> On Sun, Feb 26, 2017 at 1:15 PM, sebb <sebbaz@gmail.com> wrote:
>
>> On 26 February 2017 at 09:52, Philippe Mouawad
>> <philippe.mouawad@gmail.com> wrote:
>> > Hello,
>> > This element might be more useful if we can make some enhancements to it:
>> >
>> >    - First rename it as its name is far from clear. Something like Mail
>> >    Alerter
>> >    - Make it send notifications on richer conditions:
>> >       - Percentile 90 of response time (Excluding Transaction Controller)
>> >       reaching a limit
>> >       - Errors percentage reaching a limit
>> >       - Burst in error percentage
>> >       - ...
>> >       - Possibly be able to express a rule in Groovy
>> >
>> > Currently it seems to me far from useful as is.
>>
>> I suggest you ask on the User list how it is being used.
>>
>> Also, I would be wary of adding too much functionality into the listener.
>> It should just be a means to send the email, the same as other
>> Listeners are a means to displaying results or recording them in a
>> file or database.
>> Error checks should be done by Assertions.
>>
> In my understanding Assertion applies on every sample. Can it apply to a
> history of samplers ?

Listeners also apply to each sample, yet the Mailer Visualizer is able
to look at history, so Assertions can use the same technique.

> Maybe a simple addition would be to allow Mailer Visualizer to send a mail
> based on variable or property value (true/ false).
> This would make it very flexible. We could this way combine it with
> GenerateSummary Result or a dedicated "Alert" listener which could set this
> variable or property.

That would be better, but the "Alert" test element should be an
Assertion not a Listener.

This would allow it to be more general; e.g. one could stop a test if
the error percentage was too high.

>
>> As far as possible, each test element should have a single role.
>> This reduces duplication and allows easier testing.
>>
> I agree
>
>>
>> >
>> > Regards
>> >
>> > Philippe
>>
>
>
>
> --
> Cordialement.
> Philippe Mouawad.

Mime
View raw message