Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

HTTP request spewer / load generator

branch: master

Fetching latest commit…

Octocat-spinner-32-eaf2f5

Cannot retrieve the latest commit at this time

Octocat-spinner-32 include
Octocat-spinner-32 src
Octocat-spinner-32 .gitignore
Octocat-spinner-32 COPYING
Octocat-spinner-32 INSTALL
Octocat-spinner-32 Makefile.am
Octocat-spinner-32 README.rst
Octocat-spinner-32 WIP
Octocat-spinner-32 autogen.sh
Octocat-spinner-32 configure.ac
README.rst

README for spew

Spew is a HTTP request generator/spewer.

It's written to be fast and straight forward. It currently does no measurements and is pretty mean. But it's also fast.

It has the following features:

  • Generate HTTP requests to a server of your choice
  • Manipulate the Host-header
  • Tries to keep -p conns=X open connections at any given time
  • Sends -p reqs=X requests over each connection - IN ONE GO.
  • Ignores replies.
  • It's fast. Or more accurately: It does nothing.
  • It's single-threaded.
  • Native IPv6 support (it was easier than not to have it!)

As an example of speed, I was able to do 260-290 thousand requests per second against Varnish with spew using less than 10% cpu on my workstation at home (granted, that's a fast computer). On my laptop (i5 M520, 2.4GHz), spew easily does 30k-50k req/s with 3k-5k conn/s against a Varnish-server on localhost.

Version history

  • 0.4 - Adds -p rand=<int> option to provide a random factor to urls. Minor build tweaks.
  • 0.3 - Initial released version, sort of. 0.1, 0.2 and 0.3 were released on the same day.

Tips

  • If you want to test more than 1000-ish connections, you need to increase the ulimit for number of file descriptors.

  • Spew is relatively slow at opening connections. Or rather, opening connections is a slow task in general.

  • If you open connections too fast, you may want to set:

    sysctl net.ipv4.tcp_tw_recycle=1
    

    Otherwise you'll quickly run out of local ports.

  • Use spew -p url=/ -p otherparam=whatever -h param > ~/.config/spew to store your default options.

  • Run multiple spew-processes if you starve a single CPU thread. Let me know if this ever speeds things up for you!

Installation and usage

Dependencies:

  • C compiler
  • make
  • awk
  • basic build stuff
  • Linux (because of epoll)

On ubuntu 12.04:

apt-get install build-essential automake tcl

From a repo: ./autogen.sh && ./configure && make && make install From a tar-ball: ./configure && make && make install

Running: spew --help

If it's not in --help, it's not in spew. If it is, then someone got drunk, wohoo!

See WIP (Work In Progress) for day-to-day changes. The content of WIP as it was upon compile-time is usually printed upon startup.

Thanks

History

The core part of spew, src/http.c, was hacked together in about 10 hours total by Kristian Lyngstøl for Varnish Software.

Everything else was then later borrowed from an old defunct project I had that only did parameter-parsin, config file support and all the typical boring stuff. That's why http.c and the rest is pretty different. If you find references to X-related stuff, let me know. They are old.

Something went wrong with that request. Please try again.