3
votes

I've successfully set up a test plan in Jmeter that can execute 500 threads and loop through a few http requests on a single machine and integrated with Hudson. So let's say I want to increase my threads to 6000.

For starters I set up a few machines to accommodate a smaller total thread number like 1500. I use the built-in feature of running Jmeter in distribution mode. I enter 1500 threads in the master test plan. So here are my $10 questions:

  1. Why does each of the client run 1500 threads and not intelligently share the total thread count from the controller/master? (Did I configure my test plan incorrectly?)
  2. Is this even possible for Jmeter to also keep track of each machine's number of threads allocated to them and also keep track of the consecutive counting of counters?

Did I totally misunderstand the concept of executing load in distribution mode (I think so)? Do I manually need to configure variables, threads, loops etc. and run them separately and manually on each machine if I want to split the load?

Thanks for reading!

1

1 Answers

3
votes
  1. This is not how it works, distributed testing has the copy of test plan and sends it to jmeter-servers so if you want to have a total of 1500, set number in plan to 1500 divided by number of jmeter-server
  2. Each jmeter-server tracks its own number but does not report it to Controller. But you will have this in output report

Read this:

With JMeter what you usually have to replicate in this mode is:

  • data used by CSV

  • jmeter.properties and user.properties

But for test plan you only need to have it on Controller.