Grizzly is a framework to be able to easily define load scenarios, and is primarily built on-top of two other frameworks.
Locust: Define user behaviour with Python code, and swarm your system with millions of simultaneous users.
Behave: Uses tests written in a natural language style, backed up by Python code.
behave is abused for being able to define locust load test scenarios using gherkin. A feature can contain more than one scenario and all scenarios will run in parallell. This makes it possible to implement load test scenarios without knowing python or how to use locust.
Locust are a group of certain species of short-horned grasshoppers in the family Arcididae that have a swarming phase.
The name grizzly was chosen based on the grasshopper Melanoplus punctulatus, also known as grizzly spur-throat grasshopper. This species prefers living in trees over grass, which is a hint to Biometria1, where grizzly originally was created.
1 Biometria is a member owned and central actor within the swedish forestry that performs unbiased measurement of lumber flowing between forest and industry so that all of Swedens forest owners can feel confident selling their lumber.
More detailed documentation can be found here and the easiest way to get started is to check out the example.
A number of features that we thought locust was missing out-of-the-box has been implemented in grizzly.
Support for synchronous handling of test data (variables). This is extra important when running locust distributed and there is a need for each worker and user to have unique test data, that cannot be re-used.
The solution is heavily inspired by Karol Brejnas locust experiments - feeding the locust. A producer is running on the master (or local) node and keeps track of what has been sent to the consumer running on a worker (or local) node. The two communicates over a dedicated ZeroMQ connection.
When the consumer wants new test data, it sends a message to the server that it is available and for which scenario it is going to run. The producer then responds with unique test data that can be used.
Listeners for both InfluxDB and Azure Application Insights are included. The later is more or less appinsights_listener.py, from the good guys at Svenska Spel, but with typing.
They are useful when history of test runs is needed, or when wanting to correlate load tests with other events in the targeted environment.
locust comes with a simple user for loading an HTTP(S) endpoint and due to the nature of how the integration between behave and locust works in grizzly, it is not possible to directly use locust.user.users provided users, even for HTTP(S) targets.
RestApiUser: send requests to REST API endpoinds, supports authentication with username+password or client secretServiceBusUser: send to and receive from Azure Service Bus queues and topicsMessageQueueUser: send and receive from IBM MQ queuesBlobStorageUser: send and receive files to Azure Blob StorageIotHubUser: send/put files to Azure IoT Hub
All failed requests are logged to a file which includes both header and body, both for request and response.
pip3 install grizzly-loadtester
pip3 install grizzly-loadtester-cliDo not forget to try the example which also serves as a boilerplate scenario project, or create a new grizzly project with:
grizzly-cli init my-grizzly-projectThe easiest way to start contributing to this project is to have Visual Studio Code (with "Remote - Containers" extension) and docker installed. The project comes with a devcontainer, which encapsulates everything needed for a development environment.
It is also possible to use a python virtual environment, but then you would have to manually download and install IBM MQ libraries, and install grizzly dependencies.
sudo mkdir /opt/mqm && cd /opt/mqm && wget https://ibm.biz/IBM-MQC-Redist-LinuxX64targz -O - | tar xzf -
export LD_LIBRARY_PATH="/opt/mqm/lib64:${LD_LIBRARY_PATH}"
cd ~/
git clone https://github.com/Biometria-se/grizzly.git
cd grizzly/
python -m venv .venv
source .venv/bin/activate
python -m pip install -e .[dev,ci,mq,docs]