ดูต้นฉบับสำหรับ Lightweight Web Load Testing Tool with Curl and Ruby

This is a simple web load testing tool using Curl.  It is written in Ruby, but you don't have to know much Ruby to use it.

First you model your visitors in some file, e.g., <tt>load_test.rb</tt>:

<pre>
visitor "Viewer" do       # The visitor type name should be Capitalized, 
                          # and follow the Ruby constant naming rule, 
                          # because it shall be used to create a class on that name
  stores_cookies

  site_url "http://mysite"

  get "/"
  post "/login", {:login => 'guest', :password => 'hello-world'}
  post "/list"
  get "/"
end

visitor "Poster" do
  stores_cookies

  site_url "http://mysite"

  get "/"
  post "/login", {:login => 'user', :password => 'hello'}
  post "/list"
  post "/submit", {
    :image => {:type => :file, :data => 'picture.png'},
    'category' => 58
  }, :multipart => true
end
</pre>

Then you call
 ruby runner.rb load_test Poster 10 Viewer 20 -t 60
which loads your definition, creates 10 Poster and 20 Viewer, and executes them for 60 seconds.  After it runs, it'll run and give you some report.

==Want to try?==

'''What do you need:''' you must have [http://www.ruby-lang.org/ ruby] and [http://curl.haxx.se/ curl] (command line) installed on your system.

'''Download:''' Just go to its [http://github.com/jittat/curlby-loadtest/tree/master github repository].

'''How to install?''' You don't install.  There are two files, just place it anywhere you like.

==Documents==
After you untar the package and place it somewhere, you have to specify how each type of visitors uses you web.  Then you just call the runner.

===How to specify visitor types===
You write that as a simple ruby code.

You place each type of visitor in a <tt>visitor</tt> block:
 visitor ''"YourVisitorType"'' do
   ''# you will write some code here
 end

You can have many visitor types.

Technically, this block will create a class named ''YourVisitorType''; therefore, as a ruby naming rule, it has to be capitalized and does not contain any weird characters.

Inside the block you can place the following options:
* '''<tt>site_url</tt>''' ''base_url''  
:: This ''base_url'' will be used to prefix every http request.
* '''<tt>stores_cookies</tt>'''  
:: Use this to enable cookies.

It's time for actions.  
* '''<tt>get</tt>'''
: Formally, it is a class method with the following signature
      get(''url'',''params''=nil)
: You pass the ''url'' as a string, and the GET parameters as a ruby hash.  The ''url'' is prefixed with the <tt>site_url</tt>.  You have to specify the whole url if you do not provide the <tt>site_url</tt>.  The parameter is optional.  For example if you want to get from <tt>"/view"</tt> with parameter <tt>page=10</tt>, you write
      get "/view", {"page" => 10}
: You can also use symbols as a key, e.g.,
      get "/view", {:page => 10}

* '''<tt>post</tt>'''
: This method signature is
      post(''url'',''params''=nil,''options''=nil)
: <tt>''params''</tt> and <tt>''options''</tt> are ruby hash.  The general idea of ''params'' is pretty much the same as in <tt>get</tt>, except when you want to post a file.  In that case, your data is also a hash with two keys: <tt>:type</tt> and <tt>:data</tt>; you set <tt>:type</tt> to <tt>:file</tt>, and <tt>:data</tt> to your posting file name.  Furthermore, to post a file you have to set the <tt>:multipart</tt> option to <tt>true</tt>, which is the only option for now. See the example above.

You might notice that when calling <tt>get</tt> and <tt>post</tt> you do not have to put their arguments in parentheses: this is because Ruby does not require that.   Also, you may also notice that when setting the <tt>:multipart</tt> option to <tt>true</tt>, you do not have to put them in a hash (i.e., in curly brackets); this is, again, done automatically by Ruby.  The last line from the above example can be written as
        post("/submit", {
          :image => {:type => :file, :data => 'picture.png'},
          'category' => 58
        }, {:multipart => true})
as well, but it looks kind of ugly.

====visitor's id substitution====
Each visitor object has an auto-generated id, starting from 1, which can be used when describing its operations.  Currently, the feature is minimal.

On the post parameter values or the post parameter values, if you put <tt>${id}</tt>, it will be substitute with the visitor's id.  If you want the string <tt>${id}</tt> in your parameters, put in <tt>$${id}</tt>.  For example,

 "user${id}"     # => "user10"
 "user$${id}"    # => "user${id}"

===Calling the runner===
The runner can be called with
 ruby runner.rb ''arguments''
It usage reads.
 using: ruby runner.rb <visitor_file> [<type> <number>] [<type> <number>] ... [options]
   * visitor_file : your visitor definition file, (with or without .rb)
   * type, number : the type and the number of visitors of that type
   * options      : any of the following
                      -t <sec>   specify how long (in seconds)
                      -d         dry-run: run, but make no real http requests

: ''I'll add more document later''

==Releases==
Current release is 0.03.  The following is the release note.

 Release 0.03
 ============
 Date: 3 apr 2008
 Problems:
 * If some curl process goes to sleep mode and never comes back, the
   runner will wait forver.
 Changes:
 * Add some work around for the above problem, by intercepting
   Interrupt exception (when user presses c-break) during thread.join.
   The program still produces statistics

===Known problems===
* When running a large number of threads (say 100), the <tt>cURL</tt> processes may go to sleep mode and never wake up.  In this case runner will wait forever.

==Suggestions and bug report==
You can leave your comments at the [[Talk:Lightweight Web Load Testing Tool with Curl and Ruby|talk page]].  Or you can send e-mail to me at gmail. My account name is jittat.