mod_qos

Additional Information

 

This page gives you additional information on certain aspects of mod_qos to get a better understanding about how to use the module.



Directives

The module is configured by directives. All directives process the connection, HTTP request, and response data in a pre-defined sequence. The following graph shows the order in which the directives work.

directive sequence


Rules

mod_qos allows you to configure different kind of rules. The main component of a rule is its counter. A rule measures either the concurrency (how many times something happens at the same time), the occurrence (how often does something happen in a certain amount of time), or the throughput (sent amount of data or number of request) and stores this information within that counter.

Every rule has it's own threshold and maintains its own counter. A rule is identified by either an URL pattern/matching string or by an environment variable name. You can configure as many rules as you want.

Rule

Note: Some counters are only available once. This applies to the counters of the rules using the QS_Block, QS_SrvSerialize, and QS_Serialize environment variables.


Environment Variables

The Apache web server provides a mechanism for storing information in so called environment variables. mod_qos uses these variables to exchange data respectively signalize events between different rules defined by the corresponding directive. These variables can also be written or read by other Apache modules, such as mod_setenvifplus or mod_setenvif .

Example:
The SetEnvIf directive is used to set the LimitLogin variable if the request line matches the ^/wp-login.php pattern while the QS_ClientEventLimitCount directives increments the repeat counter having the same name if the variable is present.
Event set by SetEnvIf and processed by QS_ClientEventLimitCount

It is also possible to write the values of these variables to your log file using the format string %{VARNAME}e within the TransferLog/CustomLog directives. Or you can use them within error pages using server-side includes (SSI).

Note: Whenever you use a directive (such as SetEnvIfPlus ) which can either process request attributes (such as HTTP headers) or environment variables, you must make sure that a client can not bypass your rules by sending a request header with the same name as the environment variable used in your configuration. Use either the request header filter or the QS_UnsetReqHeader directive to prevent anyone from sending a request header with the same name as the variable you have defined.


Concurrency Counter

A "concurrency counter" is used to determine how many times something happens at the same time (concurrent), e.g. HTTP requests accessing the same resource/URL at the same time. The rules using this counter type are either defined by an environment variable name or an URL pattern (regular expression or a string matching the request's URL). Such a rule automatically increments the counter when the Apache web server starts to process a matching request and decrements the counter when the request processing is completed.

You have to configure a threshold and the rule's variable name resp. URL pattern. Requests (or new connections) are denied as soon as the configured threshold is reached.

Directives using this counter type are:

Also the QS_SrvMaxConn, QS_SrvMaxConnClose, and QS_SrvMaxConnPerIP directives use this counter type, although with fewer parameter options.

Sample Use Case

Now let us look at an example to show where these rules can be used. Let's assume that you have two applications. We call them "A" and "B". Application "A" has been deployed on path /app/a and "B" on /app/b.
     As long as the server has enough resources, users can access both applications the same time without influencing each other.
All requests are processed quickly and the workers (w) become free again to serve new requests.

  trouble-free
     But if the application "B" becomes slow, the duration of request processing increases and all workers (w) become busy. There are no free workers left and application "A" becomes unavailable because of that.

disruption
     A QS_LocRequestLimit or QS_LocRequestLimitMatch rule can help in such a situation. mod_qos limits the number of requests to application "B" so that application "A" remains available even if application "B" has performance problems. disruption

Such a scenario can occur due to various infrastructure problems, e.g., by slow database queries.
A similar situation can also arise through an external influence: if someone penetrates application "B" with a HTTP GET / POST flood DoS attack, then application "A" could also become unreachable. A QS_LocRequestLimit rule can prevent this.


Repeat Counter

"Repeat counters" limit the number how often (Cr) something is allowed to happen in a certain amount of time (Td). These rules trigger a timer whenever the defined event occurs the first time and start to count every subsequent event until the timer expires. If the event counter reaches the defined limitation, requests are blocked until the time is up.

reapet counter

All repeat counters allow you to define an event which shall increment the counter if they occur. You also have to configure a duration Td and the threshold Cr, defining how many events are allowed within the time Td.

Directive parameter example:
QS_ClientEventLimitCount

While the counter is automatically cleared (set to 0) when the time Td is up, you might also configure additional events to decrement or clear the counter earlier.

The directives using this counter type are:

Note: Some directives support the increase of the counter by more than "1" taking the variable's value into account.


Throughput Control

Throughput control is implemented by measuring the current usage and calculating a necessary delay which needs to be applied to the data processing in order to achieve the desired limitation (closed loop control system).

closed loop

mod_qos can limit the bandwidth when downloading data from your web server to the client. This throughput control can be configured by the following directives:

Note: Throughput control should always be used with a concurrency counter because the added delay increases the number of simultaneous requests.

Requests per Second

It is also possible to limit the number or requests per second to a resource. This control function is less accurate than the bandwidth limitation, since the measurement of the request rate takes longer (several seconds) and the request delay is more coarse-grained.
The following directive can be used to limit the number of requests per second:


Serialization

mod_qos offers you the option to serialize requests. Serialization means, that requests are processed one after the other. Incoming requests are queued if another request is in process and have to wait until the previous request is finished.

serialization

Requests, which shall be serialized, are tagged by one of the following environment variables:

Serialization might be applied on a per server level (serializing all HTTP requests) or on a per client level (serializing multiple requests coming from the same client/IP address).


Error Pages and Server Side Includes (SSI)

Custom error documents to be used by mod_qos are either configured using the QS_ErrorPage directive or the QS_ErrorPage variable.
You may also use Apache's server-side includes (SSI) to generate the content of the error document dynamically. The error codes and other variables set by mod_qos can be used.

Sample configuration:
AddType                   text/html .shtml
AddOutputFilter           INCLUDES .shtml
QS_ErrorPage              /errorpages/qs_error.shtml

Sample page:
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <!-- 
      --  mod_qos sample SSI error page (Apache 2.4)
      -->
    <meta http-equiv="Cache-Control" content="no-cache, no-store"/>
    <meta http-equiv="expires" content="0" />
    <title>ERROR - <!--#echo var="REDIRECT_ERROR_NOTES" --></title>
  </head>
  <body>
  <p>
    <b><i>sorry - the server was unable to complete your request</i></b>
  </p>
  <p>
        code: mod_qos(<!--#echo var="QS_ErrorNotes" -->)<br>
    <!--#if expr="v('REDIRECT_ERROR_NOTES') =~ /00[0-9]/" -->
        reason: initialisation failure
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /0[18][0-9]/" -->
        reason: request rule
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /03[0-9]/" -->
        reason: connection rule
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /[01]4[0-9]/" -->
        reason: request filter
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /05[0-9]/" -->
        reason: bandwidth limitation
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /[01]6[0-9]/" -->
        reason: client limitation <br>
        remaining time: <span id="remaining"><!--#echo var="QS_Limit_Remaining" --></span> seconds
        <script type="text/javascript">
        <!--
        setInterval(function () {
          var msg = document.getElementById('remaining');
          if(msg) {
            var value = msg.innerHTML;
            var remainTime = value - 1;
            if(remainTime < 0) {
              window.location = window.location.pathname;
            } else {
              msg.innerHTML = remainTime;
            }
          }
        }, 1000);
        //-->
        </script>
    <!--#elif expr="v('REDIRECT_ERROR_NOTES') =~ /10[0-9]/" -->
        reason: GEO location limitation
    <!--#else -->
        reason: generic failure
    <!--#endif -->
  </p>
  </body>
</html>


User Tracking

It might be necessary to identify individual users to define appropriate QoS rules. For this reason, mod_qos can set a cookie containing a unique identifier. This identifier is then written to the mod_qos_user_id environment variable and you can add it to your log files by the format string %{mod_qos_user_id}e. This allows you to identify all requests issued by a user.

This feature is enabled by the following directive:

  • QS_UserTrackingCookieName <name> [<path>] [<domain>] ['session'] ['jsredirect']
    The parameter "name" defines the cookie's name and "domain" (optional) the domain attribute for the Set-Cookie header. The string "session" can be defined if the cookie should not be stored by the session (but can be deleted when the user closes this browser).
You can also enforce the browser to support cookies by specifying the "path" parameter for the QS_UserTrackingCookieName directive. This parameter defines an error document and mod_qos answers the request with a redirect (302) to this document when setting the cookie initially. The browser will follow the redirect and mod_qos redirects the browser back to the initially requested page if the request to this error document contains the tracking cookie. If the browser did not send the cookie, the error document is shown.

user tracking cookie enforcement

Note: You can exclude certain clients from this enforcement by setting the DISABLE_UTC_ENFORCEMENT environment variable at server level (outside Location), e.g., to allow crawlers not supporting cookies to access your site.

Sample configuration:
QS_UserTrackingCookieName qstrack /errorpages/cookiecheck.html session
QS_SessionKey             sB.F4_0%D700ahXT2

Sample page:
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta http-equiv="Cache-Control" content="no-cache, no-store"/>
    <meta http-equiv="expires" content="0" />
    <title>Cookie Check Page</title>
  </head>
  <body>
    <h1>Cookie Check Failed</h1>
     <p>
       Please enable cookies in your browser.<br>
       Bitte schalten Sie in Ihrem Browser Cookies ein.<br>
       Activez les cookies dans votre navigateur, s'il vous pla&icirc;t.<br>
       Por favor active las cookies en su navegador.<br>
       Si prega di attivare i cookies nel proprio browser.<br>
     </p>
  </body>
</html>


Request Statistics Using qslog

qslog is a command line tool to generate usage statistics data from log files. It can generate the data in real time if defined within your Apache's server configuration.
Sample configuration:
CustomLog "|/usr/bin/qslog -o logs/qslog.csv -x -f ISBDk" "%h %>s %b %D %k"

Or you can use it to process any existing log file using the post processing option -p. This does not only work for Apache log files but for any other log file as well. You just have to know what's in the log and configure the format string argument -f of the qslog command accordingly.

Example:
log format definition

The output shows the data as a function of time: a summary of what happened every minute. Each line includes all measured values as semicolon spearated name/value pairs (CSV).

You can use the spreadsheet program of your choice to process the output:

spreadsheet



© 2024, Pascal Buchbinder