README.md

Load Testing with Locust

Assume API performance is one of your concerns, you may want to use some tools to run load testing against your APIs. In our scenario, we use Locust to run load testing against API Management. Locust is an open source load testing tool and you write your load tests in Python. In the following sections, we will first introduce how to write Locust tests and then we will show you how to run Locust tests locally and remotely via AKS.

Writing Locust tests

To write a Locust test, you need to define two classes:

• A subclass of TaskSet
• A subclass of Locust. In our case, we use FastHttpLocust for better performance.

Let's take a look at load test against Echo API as example. When you create API Management, Echo API is created by default as a simple API example.

def Echo(self):
    return self.client.get(
        "echo/resource?param1=sample", 
        headers = {
            "Ocp-Apim-Subscription-Key": Variables.sub_key
        },
        catch_response = True
    )

class EchoTaskSet(TaskSet):
    @task
    def EchoTask(self):
        with Echo(self) as response:
            if response.status_code != 200:
                response.failure(f"Code: {response.status_code} | Content: {response.text} | Task: EchoTask")

class EchoLocust(FastHttpLocust):
    wait_time = Variables.wait_time
    task_set = EchoTaskSet

There are three parts in the above code snippet.

• Echo function calls the Echo API in API Management and returns the response.
• EchoTaskSet class defines the task set. It calls the Echo function and checks the response. It will record the error if the returned HTTP status code is not 200.
• EchoLocust class defines what Locust will attack. In this case, all Locusts will do the EchoTaskSet and they will wait for wait_time seconds between two tasks. Usually the wait_time can be randomed in a range.

However, Echo API is not our target. In API Management, we implemented Patient API with several API operations e.g., patient creation, patient search, etc. and they are our load test target. So in our case we've written several Locust tests and some important files are listed below.

• variables.py defines some variables like wait_time, sub_key (API Management subscription key) and so on.
• locustfile.py defines all TaskSet and Locust classes. We defined one Locust class for each TaskSet which enables users to run each TaskSet separately. In addition, there is one Locust class called AllLocust which contains all TaskSet.
• tasks.py defines the Locust class you want to use to run the load test. For example, use from locustscripts.locustfile import AllLocust if you want to test all APIs. If you just want to test patient creation, then just use from locustscripts.locustfile import CreatePatientLocust instead.

Running Locust tests locally

After writing Locust tests, you can first run those tests locally to check if your Locust tests work or not.

Prerequisites

Install Locust with pip:

pip install locustio

How to run tests locally

1. Assign your API Management subscription key to sub_key in variables.py
2. Go to LoadTest folder and run the following command.

locust -f tests/tasks.py

3. Open your browser and go to localhost:8089. You need to input three things.

• Number of total users
• Hatch rate
• Host, in our case it should be the hostname of API Management

If everything works fine, you should be able to see the similar screenshot.

Running Locust tests remotely via AKS

Although it's very easy to run Locust tests locally, your local machine may not have enough power to simulate huge amounts of users. In that case, you can run Locust tests remotely via AKS. The idea is that you deploy your Locust tests in AKS containers and attack the API Management. Since you use AKS, you have the full control of the infrastructure like how many nodes to provision, how many pods (Locust workers) you want to spin up, etc.

Prerequisites

• Bash (to run script)
• Azure CLI latest version
• kubectl, you can use az aks install-cli to install it after you install Azure CLI
• Helm 3

How to run tests via AKS

1. Assign your API Management subscription key to sub_key in variables.py
2. Assign following variables in azure_cli_deploy_and_run_tests.sh
   • LOCUST_TARGET_HOST: the hostname of API Management
   • CLIENT_ID and CLIENT_SECRET: the service principal for AKS authentication
   • LOCATION: the region for AKS and storage account
   • RESOURCE_GROUP: the resource group for AKS and storage account
3. Since all scripts are in azure_cli_deploy_and_run_tests.sh, just run this bash to run tests in AKS cluster. After running Locust tests, all resources will be deleted automatically.
4. Check the test results like sample test results

Below diagram shows how the azure_cli_deploy_and_run_tests.sh script works.

References

• Locust on GitHub
• Locust Helm Chart
• Increase Locust's performance with a faster HTTP client