1 <?xml version="1.0" encoding="ISO-8859-1"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en"><head><!--
4 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
5 This file is generated from xml source: DO NOT EDIT
6 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
8 <title>mod_unique_id - Apache HTTP Server</title>
9 <link href="../style/css/manual.css" rel="stylesheet" media="all" type="text/css" title="Main stylesheet" />
10 <link href="../style/css/manual-loose-100pc.css" rel="alternate stylesheet" media="all" type="text/css" title="No Sidebar - Default font size" />
11 <link href="../style/css/manual-print.css" rel="stylesheet" media="print" type="text/css" /><link rel="stylesheet" type="text/css" href="../style/css/prettify.css" />
12 <script src="../style/scripts/prettify.js" type="text/javascript">
15 <link href="../images/favicon.ico" rel="shortcut icon" /></head>
17 <div id="page-header">
18 <p class="menu"><a href="../mod/">Modules</a> | <a href="../mod/quickreference.html">Directives</a> | <a href="http://wiki.apache.org/httpd/FAQ">FAQ</a> | <a href="../glossary.html">Glossary</a> | <a href="../sitemap.html">Sitemap</a></p>
19 <p class="apache">Apache HTTP Server Version 2.5</p>
20 <img alt="" src="../images/feather.gif" /></div>
21 <div class="up"><a href="./"><img title="<-" alt="<-" src="../images/left.gif" /></a></div>
23 <a href="http://www.apache.org/">Apache</a> > <a href="http://httpd.apache.org/">HTTP Server</a> > <a href="http://httpd.apache.org/docs/">Documentation</a> > <a href="../">Version 2.5</a> > <a href="./">Modules</a></div>
24 <div id="page-content">
25 <div id="preamble"><h1>Apache Module mod_unique_id</h1>
27 <p><span>Available Languages: </span><a href="../en/mod/mod_unique_id.html" title="English"> en </a> |
28 <a href="../fr/mod/mod_unique_id.html" hreflang="fr" rel="alternate" title="Français"> fr </a> |
29 <a href="../ja/mod/mod_unique_id.html" hreflang="ja" rel="alternate" title="Japanese"> ja </a> |
30 <a href="../ko/mod/mod_unique_id.html" hreflang="ko" rel="alternate" title="Korean"> ko </a></p>
32 <table class="module"><tr><th><a href="module-dict.html#Description">Description:</a></th><td>Provides an environment variable with a unique
33 identifier for each request</td></tr>
34 <tr><th><a href="module-dict.html#Status">Status:</a></th><td>Extension</td></tr>
35 <tr><th><a href="module-dict.html#ModuleIdentifier">Module Identifier:</a></th><td>unique_id_module</td></tr>
36 <tr><th><a href="module-dict.html#SourceFile">Source File:</a></th><td>mod_unique_id.c</td></tr></table>
40 <p>This module provides a magic token for each request which is
41 guaranteed to be unique across "all" requests under very
42 specific conditions. The unique identifier is even unique
43 across multiple machines in a properly configured cluster of
44 machines. The environment variable <code>UNIQUE_ID</code> is
45 set to the identifier for each request. Unique identifiers are
46 useful for various reasons which are beyond the scope of this
49 <div id="quickview"><h3 class="directives">Directives</h3>
50 <p>This module provides no
54 <li><img alt="" src="../images/down.gif" /> <a href="#theory">Theory</a></li>
55 </ul><ul class="seealso"><li><a href="#comments_section">Comments</a></li></ul></div>
56 <div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
58 <h2><a name="theory" id="theory">Theory</a></h2>
61 <p>First a brief recap of how the Apache server works on Unix
62 machines. This feature currently isn't supported on Windows NT.
63 On Unix machines, Apache creates several children, the children
64 process requests one at a time. Each child can serve multiple
65 requests in its lifetime. For the purpose of this discussion,
66 the children don't share any data with each other. We'll refer
67 to the children as <dfn>httpd processes</dfn>.</p>
69 <p>Your website has one or more machines under your
70 administrative control, together we'll call them a cluster of
71 machines. Each machine can possibly run multiple instances of
72 Apache. All of these collectively are considered "the
73 universe", and with certain assumptions we'll show that in this
74 universe we can generate unique identifiers for each request,
75 without extensive communication between machines in the
78 <p>The machines in your cluster should satisfy these
79 requirements. (Even if you have only one machine you should
80 synchronize its clock with NTP.)</p>
83 <li>The machines' times are synchronized via NTP or other
84 network time protocol.</li>
86 <li>The machines' hostnames all differ, such that the module
87 can do a hostname lookup on the hostname and receive a
88 different IP address for each machine in the cluster.</li>
91 <p>As far as operating system assumptions go, we assume that
92 pids (process ids) fit in 32-bits. If the operating system uses
93 more than 32-bits for a pid, the fix is trivial but must be
94 performed in the code.</p>
96 <p>Given those assumptions, at a single point in time we can
97 identify any httpd process on any machine in the cluster from
98 all other httpd processes. The machine's IP address and the pid
99 of the httpd process are sufficient to do this. A httpd process
100 can handle multiple requests simultaneously if you use a
101 multi-threaded MPM. In order to identify threads, we use a thread
102 index Apache httpd uses internally. So in order to
103 generate unique identifiers for requests we need only
104 distinguish between different points in time.</p>
106 <p>To distinguish time we will use a Unix timestamp (seconds
107 since January 1, 1970 UTC), and a 16-bit counter. The timestamp
108 has only one second granularity, so the counter is used to
109 represent up to 65536 values during a single second. The
110 quadruple <em>( ip_addr, pid, time_stamp, counter )</em> is
111 sufficient to enumerate 65536 requests per second per httpd
112 process. There are issues however with pid reuse over time, and
113 the counter is used to alleviate this issue.</p>
115 <p>When an httpd child is created, the counter is initialized
116 with ( current microseconds divided by 10 ) modulo 65536 (this
117 formula was chosen to eliminate some variance problems with the
118 low order bits of the microsecond timers on some systems). When
119 a unique identifier is generated, the time stamp used is the
120 time the request arrived at the web server. The counter is
121 incremented every time an identifier is generated (and allowed
124 <p>The kernel generates a pid for each process as it forks the
125 process, and pids are allowed to roll over (they're 16-bits on
126 many Unixes, but newer systems have expanded to 32-bits). So
127 over time the same pid will be reused. However unless it is
128 reused within the same second, it does not destroy the
129 uniqueness of our quadruple. That is, we assume the system does
130 not spawn 65536 processes in a one second interval (it may even
131 be 32768 processes on some Unixes, but even this isn't likely
134 <p>Suppose that time repeats itself for some reason. That is,
135 suppose that the system's clock is screwed up and it revisits a
136 past time (or it is too far forward, is reset correctly, and
137 then revisits the future time). In this case we can easily show
138 that we can get pid and time stamp reuse. The choice of
139 initializer for the counter is intended to help defeat this.
140 Note that we really want a random number to initialize the
141 counter, but there aren't any readily available numbers on most
142 systems (<em>i.e.</em>, you can't use rand() because you need
143 to seed the generator, and can't seed it with the time because
144 time, at least at one second resolution, has repeated itself).
145 This is not a perfect defense.</p>
147 <p>How good a defense is it? Suppose that one of your machines
148 serves at most 500 requests per second (which is a very
149 reasonable upper bound at this writing, because systems
150 generally do more than just shovel out static files). To do
151 that it will require a number of children which depends on how
152 many concurrent clients you have. But we'll be pessimistic and
153 suppose that a single child is able to serve 500 requests per
154 second. There are 1000 possible starting counter values such
155 that two sequences of 500 requests overlap. So there is a 1.5%
156 chance that if time (at one second resolution) repeats itself
157 this child will repeat a counter value, and uniqueness will be
158 broken. This was a very pessimistic example, and with real
159 world values it's even less likely to occur. If your system is
160 such that it's still likely to occur, then perhaps you should
161 make the counter 32 bits (by editing the code).</p>
163 <p>You may be concerned about the clock being "set back" during
164 summer daylight savings. However this isn't an issue because
165 the times used here are UTC, which "always" go forward. Note
166 that x86 based Unixes may need proper configuration for this to
167 be true -- they should be configured to assume that the
168 motherboard clock is on UTC and compensate appropriately. But
169 even still, if you're running NTP then your UTC time will be
170 correct very shortly after reboot.</p>
173 <p>The <code>UNIQUE_ID</code> environment variable is
174 constructed by encoding the 144-bit (32-bit IP address, 32 bit
175 pid, 32 bit time stamp, 16 bit counter, 32 bit thread index)
177 alphabet <code>[A-Za-z0-9@-]</code> in a manner similar to MIME
178 base64 encoding, producing 24 characters. The MIME base64
179 alphabet is actually <code>[A-Za-z0-9+/]</code> however
180 <code>+</code> and <code>/</code> need to be specially encoded
181 in URLs, which makes them less desirable. All values are
182 encoded in network byte ordering so that the encoding is
183 comparable across architectures of different byte ordering. The
184 actual ordering of the encoding is: time stamp, IP address,
185 pid, counter. This ordering has a purpose, but it should be
186 emphasized that applications should not dissect the encoding.
187 Applications should treat the entire encoded
188 <code>UNIQUE_ID</code> as an opaque token, which can be
189 compared against other <code>UNIQUE_ID</code>s for equality
192 <p>The ordering was chosen such that it's possible to change
193 the encoding in the future without worrying about collision
194 with an existing database of <code>UNIQUE_ID</code>s. The new
195 encodings should also keep the time stamp as the first element,
196 and can otherwise use the same alphabet and bit length. Since
197 the time stamps are essentially an increasing sequence, it's
198 sufficient to have a <em>flag second</em> in which all machines
199 in the cluster stop serving and request, and stop using the old
200 encoding format. Afterwards they can resume requests and begin
201 issuing the new encodings.</p>
203 <p>This we believe is a relatively portable solution to this
204 problem. The identifiers
205 generated have essentially an infinite life-time because future
206 identifiers can be made longer as required. Essentially no
207 communication is required between machines in the cluster (only
208 NTP synchronization is required, which is low overhead), and no
209 communication between httpd processes is required (the
210 communication is implicit in the pid value assigned by the
211 kernel). In very specific situations the identifier can be
212 shortened, but more information needs to be assumed (for
213 example the 32-bit IP address is overkill for any site, but
214 there is no portable shorter replacement for it). </p>
217 <div class="bottomlang">
218 <p><span>Available Languages: </span><a href="../en/mod/mod_unique_id.html" title="English"> en </a> |
219 <a href="../fr/mod/mod_unique_id.html" hreflang="fr" rel="alternate" title="Français"> fr </a> |
220 <a href="../ja/mod/mod_unique_id.html" hreflang="ja" rel="alternate" title="Japanese"> ja </a> |
221 <a href="../ko/mod/mod_unique_id.html" hreflang="ko" rel="alternate" title="Korean"> ko </a></p>
222 </div><div class="top"><a href="#page-header"><img src="../images/up.gif" alt="top" /></a></div><div class="section"><h2><a id="comments_section" name="comments_section">Comments</a></h2><div class="warning"><strong>Notice:</strong><br />This is not a Q&A section. Comments placed here should be pointed towards suggestions on improving the documentation or server, and may be removed again by our moderators if they are either implemented or considered invalid/off-topic. Questions on how to manage the Apache HTTP Server should be directed at either our IRC channel, #httpd, on Freenode, or sent to our <a href="http://httpd.apache.org/lists.html">mailing lists</a>.</div>
223 <script type="text/javascript"><!--//--><![CDATA[//><!--
224 var comments_shortname = 'httpd';
225 var comments_identifier = 'http://httpd.apache.org/docs/trunk/mod/mod_unique_id.html';
227 if (w.location.hostname.toLowerCase() == "httpd.apache.org") {
228 d.write('<div id="comments_thread"><\/div>');
229 var s = d.createElement('script');
230 s.type = 'text/javascript';
232 s.src = 'https://comments.apache.org/show_comments.lua?site=' + comments_shortname + '&page=' + comments_identifier;
233 (d.getElementsByTagName('head')[0] || d.getElementsByTagName('body')[0]).appendChild(s);
236 d.write('<div id="comments_thread">Comments are disabled for this page at the moment.<\/div>');
238 })(window, document);
239 //--><!]]></script></div><div id="footer">
240 <p class="apache">Copyright 2013 The Apache Software Foundation.<br />Licensed under the <a href="http://www.apache.org/licenses/LICENSE-2.0">Apache License, Version 2.0</a>.</p>
241 <p class="menu"><a href="../mod/">Modules</a> | <a href="../mod/quickreference.html">Directives</a> | <a href="http://wiki.apache.org/httpd/FAQ">FAQ</a> | <a href="../glossary.html">Glossary</a> | <a href="../sitemap.html">Sitemap</a></p></div><script type="text/javascript"><!--//--><![CDATA[//><!--
242 if (typeof(prettyPrint) !== 'undefined') {