On 22 July 2010 12:21, Jörg Godau <[email protected]> wrote:
>> Please don't include test plans in e-mail in future. They are
>> difficult to extract, and are just wasted bandwidth for most readers.
>
> Sorry. How else can I share the test files?
As I wrote previously:
"I suggest creating a Bugzilla issue and attaching the JMX file. Redact
any sensitive information first."
Alternatively, upload the file somewhere public and provide a link.
> Is there a Jmeter forum somewhere? That might be a more modern (and easily
> searchable) way of doing this?
>
No, this is the only forum.
There are plenty of mail archives that can be searched; see the
JMeter/Jakarta web-sites.
>> Also, the test plan has no definition for the following variables:
>> path.root
>> resource.path
>
> Yes as I said I've stripped it down - just the parts where the User
> Parameters are being used were shown here.
> The path.root is definied as ${__P('contextroot',/kita)} and
> resource.path as ${__P('resource.path',/home/jmeter/data/)}
By the way, you should not quote the property names.
Use ${__P(contextroot,/kita)}
Not ${__P('contextroot',/kita)}
> We do not (currently) overwrite these from the command line, though we would
> like to in future.
Which you won't be able to do easily unless you remove the quotes.
>
>
>> > The open record step is where it fails saying with "No record with ID
>> ${aktenzeichen} found". Which means that there was no result for the
>> search.
>>
>> No, it means that the variable "aktenzeichen" was not found.
>> Which means the RE did not match
>
> Yes and the Regexp does not match because no record with the name from the
> CSV file is found by the user currently doing the search. Even though the CSV
> files are generated specifically for records that the paired user has access
> to via specific database queries.
>
>
>> There are more threads (10 by default) than there are User Parameter
>> columns, so some values will be used by multiple threads.
>
> Yes this is deliberate - we want to hit one section of the data harder than
> another, so with 6 User Parameter columns over 10 threads we hit 1..4 with 2
> threads each and 5 and 6 only with one thread (according to the information
> from your previous post).
>
OK.
>
>
>> The way you pass in the data is rather complicated.
>>
>> I suggest you create a single CSV file with all the data, and use CSV
>> Dataset to read it.
>
> There are several reasons why we cannot do this. The main one is that under
> full load conditions we run 10 JMeter-slaves each with 10 Threads. The data
> for each Slave is different.
>
You could just create 10 different files and pass the filename on the
command-line.
> It is also very inefficient to have huge CSV files that contain
> UserID,SearchName where UserID would be the same for thousands and thousands
> of records.
>
In that case having a separate file for the search names is reasonable.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
