[They were just suggestions...] Large JMX files are anyway not an ideal solution, as the entire file is held in memory. Also, there may be multiple versions of each test element - for example each sampler is cloned per thread.
== Have you considered creating Samplers with variable data fields? You can then pick up the variable data from a file using a function such as StringFromFile or CSVRead. If a particular field cannot be varied, you might need to create multiple samplers, and put them under a set of If controllers. This works very well for us in large volume tests (up to 100k+ requests), though admittedly we tend to exercise only a few different samplers in a given test (e.g. login, repeat a short request sequence multiple times, logout) You might be able to parameterise at least some of your tests this way. The disadvantage is that the test plan will have to be created by hand, at least initially. S. -----Original Message----- From: Felix van Hove [mailto:[EMAIL PROTECTED] Sent: 24 June 2004 10:37 To: JMeter Users List Subject: Re: importing linklists Hello, I don't think, scripting is a good choice. Last year I used Tcl scripts for creating JMX, this year I'd have to write new scripts. File format changes. I'm neither convinced of the "evil suggestion". It's not intuitive. It's reusing features build for other purposes. It can't reconstruct the change between HTTP and HTTPS. And and and... If I find time, I'll do my homework, present a ready patch and ask again. Felix BAZLEY, Sebastian wrote: > Or - continuing the theme - use a scripting language to create the JMX file. > > This becomes a lot easier with the work that Mike has done on using Xstream > for parsing the file, as the JMX file format is now a lot simpler. > > S. > -----Original Message----- > From: Michael Stover [mailto:[EMAIL PROTECTED] > Sent: 23 June 2004 15:26 > To: JMeter Users List > Subject: Re: importing linklists > > > Evil suggestion: > > Use the current AccessLogSampler on the list of urls (may need a new > parser implementation, depending on format of list). > > Run two instances of JMeter. One instance has a recording proxy server > on. The other instance is started up to USE that proxy server. This > second instance runs the AccessLogSampler against the file. The samples > will then be recorded in the JMeter running the proxy server. > > -Mike > > On Wed, 2004-06-23 at 10:09, Peter Lin wrote: > >>ok, I think I understand now. To paraphrase. >> >>1. import in URL via some mechanism >>2. manually change the threadgroup to simulate user session >>3. run test >> >> >>it sounds like you would like a better solution to 1 and 2 using some >>sort of import. correct me if I'm wrong. the feature you are looking >>for is something like this. >> >>1. a file of URL's with multiple sets of request parameters >>2. use a new plugin to import the file >>3. the plugin generates a thread group with n HttpSamplers matching >>exactly the same number of parameter sets in each line. Like say each >>line does login, search and logout. the import would create three >>HttpSamplers with the correct parameters in each. >>4. run your test plan. >> >>Is that an accurate interpretation? >> >>peter >> >> >>On Wed, 23 Jun 2004 15:54:53 +0200, Felix van Hove >><[EMAIL PROTECTED]> wrote: >> >>>Hello Peter, >>> >>>thanks for the reply. I definitly should have a closer look at the > > access log > >>>sampler. Let me explain further, what I would like to get by an import > > feature. > >>>If I get individual samplers by importing a list of (HTTP-) addresses, I > > can > >>>configure requests much more precisely. In my case I need to simulate > > half a > >>>dozen different users, who are subscribing to newsletters, login, > > logout, > >>>ordering products, registering etc. I need to url-rewrite nearly every > > request, > >>>iterating over productlists, address-lists, accounts etc. The import > > itself is > >>>only the first step in building the test plan for me. >>> >>>If I am not focussing on server logfiles, I may also switch between > > different > >>>protocols, samplers, even server instances. I had the idea of stress > > testing a > >>>server by a combination of webservices- and user-requests; first I'd > > have to > >>>import SOAP requests and the usual HTTP requests, afterwards combining > > them to a > >>>complex test plan. Ok, I'm not sure, if such a scenario makes sense at > > all;-) > >>>And last but not least a practical reason: As a customer, whishing > > someone to > >>>check my site, I would copy-paste URLs from the address line of my > > browser to a > >>>file. I actually got my linklist as a power point presentation :-) >>> >>>Felix >>> >>> >>> >>>Peter Lin wrote: >>> >>>>Mike recently made a lot of nice changes to AccessLogSampler. >>>>Probably the easiest way is to write a parser that implements the >>>>parser interface. then you can simply drop the jar file into /lib/ >>>>directory to use it with the AccessLogSampler. >>>> >>>>the original idea behind the accesslog sampler was to use production >>>>access logs for simulating loads and running simulations. This is >>>>particularly useful if there's a bug that appears under very specific >>>>situations. Say you discover a bug in a webapp, so you stress test it. >>>>The bug doesn't appear with randomly generated requests. >>>> >>>>in cases like these from my own experience, it is the result of usage >>>>patterns, which random requests can not reproduce. the accesslog also >>>>lets me compare one system to another and make apples-to-apples >>>>comparison. well as close to apples-to-apples as possible. if you >>>>need help implementing the parser interface, just post a message and >>>>I'll try to explain it to the best of my ability. >>>> >>>>peter >>>> >>>> >>>>On Wed, 23 Jun 2004 12:53:08 +0200, Felix van Hove >>>><[EMAIL PROTECTED]> wrote: >>>> >>>> >>>>>I just started to prepare a jmeter add-on to import request lists, >>>>>using each request parsed from the list to create a sampler suiting >>>>>the protocol. But may be I'm overlooking an allready existing feature? >>>>>(As far as I can see, the access log parser has a much more >>>>>specialised - and different - idea.) >>>>> >>>>>Until now I'm focussing on linklists / clickstreams. The import parses >>>>>a list of links (seperated by linefeeds) and creates a HTTP sampler >>>>>(as a child of the current node) for each of them. It's taking care >>>>>for preserving the protocol (which only can be HTTP and HTTPS until >>>>>now), server name, port number, path and parameters (using >>>>>HtmlParsingUtils.createUrlFromAnchor()). >>>>> >>>>>Background: I had to build a test plan by 79 requests, the clickstream >>>>>file has 13792 characters, http parameters everywhere. Especially to >>>>>paste parameters is extremly error-prone. >>>>> >>>>>I inserted this import function as a menu point in the file menu, - >>>>>it's just opening a file-open dialog. >>>>> >>>>>Is there allready a better way for doing this? >>>>> >>>>>-- ___________________________________________________________________________ This e-mail and the documents attached are confidential and intended solely for the addressee; it may also be privileged. If you receive this e-mail in error, please notify the sender immediately and destroy it. As its integrity cannot be secured on the Internet, the Atos Origin group liability cannot be triggered for the message content. Although the sender endeavours to maintain a computer virus-free network, the sender does not warrant that this transmission is virus-free and will not be liable for any damages resulting from any virus transmitted. ___________________________________________________________________________ --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
