HPI SchulCloud load tests

To test the performance of the HPI-SchulCloud application, multiple useres will be simulated by the loadtest. It creates a provided number of users, which are different by their user-type (admin, techer, pupil), and swarms the system simultaneausly, simulating their behavior. Courses, BBB-rooms, tasks etc. will be created, edited and deleted to generate additional traffic, and is captured on the Locust web-interface for subsequent analysis. Actually the loadtest implementation provides two kinds of starting the tests:

• scalable load tests which are designed to run with a specified number of user from the locust web interface
• functional tests which do not scale but execute the same tests as the load tests, this startup typ could be used for smoke testing a fresh deployed instance or regular verfying the proper funtion of a running instance The functional tests provide the number of executed and failed tests as Prometheus metrics to allow alerting on failed tests.

Requirements

• Python (>= 3.6.10)
• ChromeDriver (>= 90.0.4430.24, just necessary for BBB-Loadtest, Document-Loadtest)
• Docker (>= 19.03.5, optional)

Preparations

1. Download ChromeDriver

• The chromedriver.exe needs to be in the same path as the locustfile.py-File

2. Configure the environment

• Shared Variables
  • PYTHONPATH=./src:${PYTHONPATH}
  • ADMIN_EMAIL=[email protected]
  • ADMIN_PASSWORD=<admin_password>
  • TEACHER_EMAIL=[email protected]
  • TEACHER_PASSWORD=<teacher_password>
  • PUPIL_EMAIL=[email protected]
  • PUPIL_PASSWORD=<pupil_password>
  • ANONYMOUS_EMAIL=[email protected]
  • ANONYMOUS_PASSWORD=<anonymous_password>
  • BBB_ROOM_COUNT : INT of Rooms of BBB
  • BBB_USER_COUNT : INT of Users per Room
  • WEIGHT_ADMIN : INT users weight
  • WEIGHT_TEACHER : INT users weight
  • WEIGHT_PUPIL : INT users weight
  • WEIGHT_ANONYMOUS : INT users weight
  • WEIGHT_ACTUAL_ANONYMOUS : INT users weight
  • WAIT_TIME_SHORT : Time in Sec
  • WAIT_TIME_LONG : Time in Sec
• Load Test specific Variables
  • FUNCTIONAL_TEST : 0
• Functional Test specific Variables
  • FUNCTIONAL_TEST : 1
  • TARGET_URL : Specify the instance to be tested as FQDN, e.g. https://nidersachsen.cloud
  • Optional:
    • TIMEINTERVAL_SEC : Time in sec to sleep between repeated tests, default is 300
    • PROMETHEUS_PORT port to expose Prometheus metrics, default is 9000

1. Install required programs with pip3 install -r requirements.txt

Run the load tests locally

Command line and web-interface

To run the load test, first open a command line and start locust with the following command:

`locust -f ./locustfile.py --host https://hackathon.hpi-schul-cloud.de --tags test`

Insert the right host after the '--host' tag*

• The '--tags test' tag is optional and will only start tasks which are marked with this tag*

Afterwards, start your webbrowser and open localhost:8089 or http://127.0.0.1:8089/. The locust web-interface should load, and you can fill out the form and start swarming.
For further information about running locust check out their documentation.

Docker

docker run -it --rm --entrypoint /bin/sh -v "$(pwd)":/app -w /app locustio/locust:0.14.4
pip install -r requirements.txt
locust -f ./locustfile.py --no-web --clients 20 --run-time 30s --host https://hackathon.hpi-schul-cloud.de --logfile $(date '+%Y-%m-%d-%H:%M:%S')-hackathon.log --csv=$(date '+%Y-%m-%d-%H:%M:%S')-hackathon

Run the functional tests locally

Execute the following command at the top level of the repository

python3 src/loadtests/functionaltests/main.py