LD-Relay Locust File

This is a locust file meant to emulate the load of launchdarkly server and mobile clients.

Setup

Prerequisites

If you are running the testing infrastructure outside of Docker you will need to install the dependencies. This project utilizes Python 3, and Locust 0.13.5. You can install the required version of Locust with:

pip3 install locustio==0.13.5

As with many Python projects you will also need to install the dependencies listed in requirements.txt. Enter the locust directory and run:

pip3 install -r requirements.txt

Optional: Configure a flag heartbeat

This will allow you gather metrics on how long it takes for a flag to propogate to clients. If you do not follow these steps, you will not get these metrics in Locust but the tests will still run.

• Create a flag with the key locust-heartbeat with the type number with two variations
• Create an api key that can modify this flag (see the example token policy below)

Modify locustfile.py

First off, take a look at launchdarkly_locust.py, this file defines the LaunchDarklyLocust and LaunchDarklyMobileLocust classes that you will be basing your locusts on.

Now open up locustfile.py. You'll see a couple of task sets and locusts that you can modify. Simply edit the task sets to emulate the behavior of your users. The example ones just initialize the client and evaluate random flags.

If you want to modify the client settings, just set the properties you'd normally pass to LDClient.Config as properties on the LaunchDarklyBasicServer and LaunchDarklyBasicMobile classes.

Feel free to create as many new task sets and locust subclasses as you want to define different classes of users and behavior. You can define what percentage of users will use each behavior using the weight property on the locust class or task sets.

For more information, read this guide

Setup your environment variables

Configure your environment variables like so:

LAUNCHDARKLY_MOBILE_KEY=mob-xxxx
LAUNCHDARKLY_SDK_KEY=sdk-xxxx
LAUNCHDARKLY_HEARTBEAT_PROJECT=my-proj-key
LAUNCHDARKLY_HEARTBEAT_API_KEY=api-xxxx
LAUNCHDARKLY_BASE_URI=http://ldrelay:8030
LAUNCHDARKLY_EVENTS_URI=http://ldrelay:8030
LAUNCHDARKLY_STREAM_URI=http://ldrelay:8030
LAUNCHDARKLY_MOBILE_KEY=http://ldrelay:8030
LAUNCHDARKLY_MOBILE_STREAM_URI=http://ldrelay:8030

You can also set the sdk_key properties on each locust class instead:

class MyLocust(LaunchDarklyMobileLocust):
  sdk_key="mob-xxx"
  # or use a different environment variable
  sdk_key=os.environ.get('SOME_OTHER_ENV')

Running the tests

• Run locust with python3 -m locust -f locustfile.py
• Open https://localhost:8089 and enter your desired users and set the host to your instance of LD-Relay

Running in docker

docker build -t locust-ld-relay .
docker run -p 8089:8089/tcp -e TARGET_URL=http://ldrelay:8030 --env-file=env locust-ld-relay:latest

Running in Distributed mode

You can test with large amounts of users by using multiple locust nodes. Simply configure the masters and slaves with the following env variables:

Master

LOCUST_MODE=master

Be sure to expose port 5557-55578 to your slave nodes

Slaves

LOCUST_MODE=slave
LOCUST_MASTER_HOST=<hostname of master>

For more information, refer to Locust Documentation

FAQ

Why is the RPS so low?

Users don't equate to requests in Locust. The number of users is the number of instances of the Locust classes defined in your locustfile. It will make as many requests as they make while running.

Additional Metrics

You can get additional metrics from LD-Relay using the prometheus support in LD-Relay along with the prometheus node exporter.

You can also import stats from Locust into prometheus using this exporter which is also available on docker hub under tarqld/locust_exporter

Locust Api Token Policy

Below is an inline policy that would allow the token you generate to update the locust-heartbeat flag in any of your projects

[
  {
    "resources": [
      "proj/*:env/*:flag/locust-heartbeat"
    ],
    "actions": [
      "updateFlagVariations"
    ],
    "effect": "allow"
  }
]