Mastering wrk: A Practical Guide to HTTP Load Testing with Lua Scripts

Introduction

wrk is an HTTP load‑testing tool that features a multithreaded design with event‑notification systems (epoll, kqueue) for high concurrency and extensibility via Lua scripts for request generation, response handling, custom reports, etc.

Integrated multithreading and event notification (epoll, kqueue) to increase concurrency.

Extensible through Lua scripts for custom request generation, response processing, reporting, and more.

Usage Example

Installation

wrk runs on Linux or Mac .

Project repository: https://github.com/wg/wrk After cloning, build the project: make The build produces a wrk executable in the current directory.

Running

Execute a test command: ./wrk -t12 -c100 -d10s http://www.baidu.com This tests baidu.com with 12 threads, 100 concurrent connections, for 10 seconds.

It is recommended not to set the thread count excessively; a good rule of thumb is 2‑4 × the number of CPU cores.

Typical output includes:

Requests/sec

– basic metric indicating how many requests are processed per second. Thread Stats – per‑thread statistics such as latency, request rate, with Avg (average), Max (maximum), and Stdev (standard deviation). The standard deviation measures data dispersion; a larger Stdev across multiple runs suggests performance variability.

Combining Scripts

Lua scripts enable custom functionality such as sending POST data, custom aggregation after all requests, or testing multiple URLs.

Example post.lua script:

-- example HTTP POST script which demonstrates setting the
-- HTTP method, body, and adding a header

wrk.method = "POST"
wrk.body   = "foo=bar&baz=quux"
wrk.headers["Content-Type"] = "application/x-www-form-urlencoded"

Run wrk with the script using the -s flag:

./wrk -t12 -c100 -d10s --script=post.lua http://www.baidu.com

wrk provides several hook functions:

setup – called once per thread after all threads are created.

init – called before each request is sent.

delay – introduces a delay between requests.

request – allows modification of the request object before it is sent.

response – processes each response.

done – called after all requests have completed.