Deploying Locust Distributed Load Testing with Docker Containers

Background Installing Locust on a Linux kernel requires many C/C++ dependencies; to reduce deployment cost, Docker containers are used to package Locust and its libraries.

Docker provides process, file‑system, and network isolation on Linux (x64) via user‑space mechanisms, enabling lightweight containers that communicate through a virtual bridge (docker0).

Solution

1. Distributed Mode – Locust runs in a master‑slave configuration; the only difference is the command‑line argument, so a single Docker image can serve both roles.

2. Image Design – The base image installs locustio and required packages. The final Locust image receives the scenario file path, mode (master or slave), target URL, and master IP via environment variables at container start, allowing dynamic configuration.

3. Dockerfile – A Dockerfile (shown in the original images) builds the Locust image with the necessary dependencies and copies the test scripts into /home/locust/.

4. run.sh Script – A custom startup script (adapted from a GitHub project) sets all parameters through environment variables, which Docker passes to the container at launch.

5. Networking – Master and slave containers communicate over port 5557 (customizable). If both run on the same host, Docker’s parent‑child container relationship allows the child to access any port of the parent. For separate hosts, the -p flag maps the master’s port to the host, enabling remote access.

6. Start Commands (same machine)

docker run --name locustmaster -v /home/like/locustDocker/src/:/home/locust/ -e LOCUST_MODE=master -e TARGET_URL=http://www.so.com -e SCENARIO_FILE=/home/locust/main.py -p 8089:8089 [image id]

docker run -v /home/like/locustDocker/src/:/home/locust/ --link locustmaster:lm -e LOCUST_MODE=slave -e TARGET_URL=http://www.so.com -e SCENARIO_FILE=/home/locust/main.py -e MASTER_HOST=172.17.0.2 --name locustclient1 -d [image id]

The --link option creates an alias ( lm) for the master container, simplifying address resolution.

7. Actual Run

System resources: 4 CPU, 4 GB RAM. Test configuration: 1 master, 4 slaves, 500 users, think time 100 ms, target URL http://www.so.com. The performance observed inside containers was comparable to running Locust directly on the host, indicating the containerized approach is viable.

8. Results – Screenshots (included in the original article) show the Locust web UI and test metrics.