2008/5/7 efj <[EMAIL PROTECTED]>:
>
> Guys, thanks for reponses.
>
> Steve, I've got your suggested solution working (single user, loop forever
> thread group, stop tread on EOF CSV input file, line for each folder).
> Thanks for the help. However, I'd rather have more than one thread all doing
> the same action concurrently. So would like to get the ForEach controller to
> work. I've tried to explain my reasoning below and also respond to Sebb's
> comments.
>
> I'm new to performance testing and am leading a project for our already
> developed client server application. I want to build up our experience and
> assets in simple stages. Thus I'm trying to do baseline testing first. I
> want to run a set of tests that will do one specific action repeatedly for
> multiple threads to get consistently reproduceable results.
>
> Rather than run tests that have various actions all happening at the same
> time, I want all users doing the same action at the same time as best
> possible. I don't want a code change for ActionA to influence the results
> for ActionB purely for the fact that the baseline tests always have the two
> actions running/hitting the web server at the same time.
>
> I want to use multiple threads to get averages from more samples (I'm
> assuming 10 different threads repeating an action 10 times is better than 1
> thread repeating the same action 100 times). Even if there is no difference,
> I also want to be able to scale up the number of threads for future
> investigations into load and capacity, so better to have tests multithreaded
> to start with.
>
> An example initial action I'm testing is a browse to a folder containing x
> objects of file y size. There'll be lots of permutations of number and size
> thus lots of different folder paths to browse to.
>
> Based on the online docs best practice I thought it better to have:
> 1) As few testplans (jmx files) as possible covering similar tests rather
> than have lots of individual testplans. That way jmeter fires up the threads
> as few times as possible and then they get on with the different tests.
> 2) Rather than the same action repeated lots of different times, use a loop
> and pass in the unique data e.g. folderpath. Hence the original post. Sebb,
> in repsonse to your question, all the threads need to be doing the same
> action to the same folder at the same time - I use a synch timer (see
> below).
>
> So my plan had been:
>
> TG (10 threads, RU 10, Loop 10)
> --Only Once
> ----Log on
> ----Test Action
> -------Synch Timer (10 threads) (all logged on threads start the action
> under test at the same time)
> --CSV data Config (has one variable called 'path', example of data in file
> 'folderA/folderBB,folderC/folderDD')
And how does this path variable relate to the browsePath variable below?
> --For Each Controller (I'm not sure on the input variable prefix? Output
> variable is 'browsePath')
> ----Loop (10)
> ------Folder Browse (using ${browsepath} I also want to include the variable
> in the sampler label so that the results can be grouped by folder name)
> ----Test Action
> ------Synch Timer (10 threads) (this way all threads complete one folder
> browse before moving onto next)
>
>
> So to clarify the questions
> 1) Is my general approach reasonable (I appreciate this can be subjective
> and influenced by many factors), so perhaps a better question - is it
> fundamentaly flawed?
It's not usual for 10 users to perform exactly the same actions at once ...
> 2) How can I set up input numbered variables that the For Each controller
> needs to use. I know all the values I want to use inadvance. Is it as simple
> as setting them up as UDVs on the Test Plan itself and avoid the CSV all
> together?! Just thought of doing that and will try it in the morning!!
Yes, that would be simpler.
> Any help with the above appreciated, especially as it's some of the basic
> concepts I'm asking about.
>
> Cheers
> --
> View this message in context:
> http://www.nabble.com/Simple-csv-input-data-iteration---all-threads-using-same-value-tp17101828p17111172.html
>
>
> Sent from the JMeter - User mailing list archive at Nabble.com.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
