Testertools Forum

1. ricky
   Member

   Reputation:44

    

   Much is left to be desired about how to effectively install and use "The Grinder" after wandering through the site (latest version 3.4) documentation and other hits I can find on the web.

   While I now have both the Console and an Agent running from command lines on a Windows (same 127.0.0.1) box, I am struggling to understand how to effectively "distribute" the Agent's grinder.properties file to the Agent "root" folder such that when it starts up it will pickup the grinder.consoleHost definition. There seems to be a chicken and and egg issue here, because you can't "distribute" files from the Console until the agent is up, but when it starts up it will only 'eat' the grinder.properties file in the folder in which the Agent starts and NOT the last one sent by the Console to the Agent in the distribution folder under the root folder.

   Anyway I don't expect to get a thread going here, but can anyone point me to a real blog that actively handles "The Grinder" questions of this nature? The SourceForge website only offers some type of mail-server thing that would like to avoid if I can (have not had much luck with these on SourceForge in the past).

   Posted 1 year ago #

2. frodon
   Member

   Reputation:110

    

   just curious what you like about Grinder, or why you are looking to use it?

   I POC'ed it several years ago (2005?) and found it confusing, but perhaps it has progressed since then. The main draw for me was the Python/Jython scripting for virtual users.

   I built my own [Python scriptable] performance test framework if you wanna have a look: http://code.google.com/p/multi-mechanize/
   (definitely not as big/robust as the Grinder)

   sorry for the hijacking your thread

   Posted 1 year ago #

3. ricky
   Member

   Reputation:44

    

   So far I am not sure that Grinder is any less confusing then when you evaluated it. The online sourceforge website is not well organized (well, from my perspective) and the documents that are presented surely make sense to the authors (who have done that writing for free) but are not well organized or easy to understand (again, at least for me). And the Java and software revision levels in the docs are way downlevel from what is available today--so the docs are old.

   And unlike a tool like LoadRunner the "Console" (equivalent to the LR Controller) does not assert control over the "Agents". Rather you start the Console and then each individual agent is given a config file (grinder.properties) that tells the agent the IP/port number of the running console it can connect to the Console. So if you have multiple agents (on multiple servers) you have to start all individually.

   And that is as far as I have managed to get. To get that far I had to google for articles that showed me how to get things started, once I downloaded the software from sourceforge. I suppose now I would find the startup steps in the sourceforge docs.

   Why am I evaluating this tool if we have LoadRunner? Because we wanted a tool that all the QA folks can use from time to time to run a very specific "multi-user" test sceneario against our product during monthly regression test. And that can be executed on any Windows test server.

   My goal (Mr. Phelps if you accept this assignment...) is to encapsulate the running of this test in a couple of batch files. Not sure if this is going to be doable, but it might be possible if we keep both the Console and a single Agent on the same PC in a path relative environment, as we only need to emulate 5 users cycling through 4 different reports for the duration of the test.

   So, anyone aware of any Grinder blogs where I ask my questions and get some accurate answers?

   The problem with this class of tool, is that the authors make no bones about it, they are tools targeted at software developers who use the same basic framework to develop their software products as the tool is developed under. So to the software savvy implied behaviors are not the issue they are for a testhead like me.

   Posted 1 year ago #

4. sublime
   Member

   Reputation:199

    

   I have recently been looking at the Grinder for possible cloud distributed load testing (Open Source) I know OpenSTA does this but I wanted to evaluate the latest versions of Jmeter and Grinder as they have both changed considerably since I last saw them.

   From what I understand of your requirements, to run a simple performance regression periodically, I don't see the need for the console. Unlike LR the number of threads (Virtual User equivalents) can be specified without a console by just using the properties file and a single agent.

   If an agent is started without a console it just executes with reference to its own properties. All the logs etc. are collected and any results can be parsed out as needed.

   Posted 1 year ago #

RSS feed for this topic

Reply

You must log in to post.