This repository provides an example of how to Load test an API using k6, Grafana and WireMock.

An example Weather API is used which makes requests to Open-Meteo, an Open Source Weather service. WireMock is used to mock Open-Meteo so that consistent Load metrics can be achieved when testing.

The example Weather API can be run using Docker. The test scripts and load test components can also be run using Docker.

• API - ASP.NET Weather API. Can be accessed using http:https://localhost:5294/swagger/index.html
• Database - MongoDB database used to store weather forecast requests
• k6 - Go based performance and automation testing framework
• WireMock - Used to mock Open-Meteo responses when load testing
• InfuxDB - Time series database used to store load test metrics
• Grafana - Used for dashboard view of load test metrics

All of the components can be run using Docker via docker-compose.

The local docker-compose profile can be used to run the API and MongoDB database:

docker-compose --profile local up -d

The load-local docker-compose profile can be used to the load testing components required for storing metrics from k6:

docker-compose --profile load-local up -d

The test.js script is used for the load test. This can be run using Docker or directly using the k6 CLI.

The steps for installing k6 can be found here: https://k6.io/docs/get-started/installation/

k6 run --out influxdb=http:https://localhost:8086/k6 -e MAX_VUS=1500 -e API_DOMAIN=http:https://localhost:5294 ./load/scripts/test.js

MAX_VUS can be changed to alter the number of concurrent users used for the load test.

The test is designed to run over 3 stages:

• Ramp-Up - Ramp up the concurrent users gradually
• Main - Run consistent load using the max concurrent users
• Ramp-Down - Ramp down the concurrent users gradually to zero

The load test involves creating a weather forecast and retrieving it using its unique ID. These steps are performed by k6 in a continuous loop and run across lots of threads based on how many concurrent users are specified. Short sleep intervals are also included to simulate a user accessing the application and waiting between actions.

The output from the Load test can be viewed using Grafana:

http:https://localhost:3000