Locust performance test template for running locust tests for a REST service*

To run the Locust scripts you need Python 3.6 or later (developed and tested with 2.7.15) with pip, and the dependences specified in requirements.txt (specific versions of: locustio, gevent and Jinja2) which can be installed by running

sudo apt install python3-pip
pip3 install -r requirements.txt

The default TEST URL http://localhost:8085 is hardcoded in the scripts for quick testing. You should always use the locust -H command line argument to override this default value and specify the base URL you want to attack with your locust swarm.

Locust tries to simulate real users, a great number of concurrent users with random behaviors including random timing. By default the simulated users wait between 0 and 500 milliseconds between requests (uniform random variable). These parameters can be changed by editing the min_wait and max_wait values at the end of each script. For maximum throughput, set both parameters to 0.

After configuring the scripts, just run them with locust. For example to run the test PT1 against the URL http://my-url:8080/ run:

locust -f locustfile-PT1.py -H http://my-url:8080/

After running this command, go to the URL printed by locust, by default http://localhost:8089/ and follow the on-screen instructions. You will be able to specify the number of clients to simulate and the ramp-up rate, and then lauch the tests and see the statistics in real time (as well as stop the tests when needed).

To run a cluster of Locust clients just add the argument --master to one of them, which will be the master, and --slave --master-host=<master's IP> to each of the slaves. The master will show the whole cluster's activity and statistics through its web interface.