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 <meta content="text/html; charset=ISO-8859-1" http-equiv="Content-Type" />
6 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
7 This file is generated from xml source: DO NOT EDIT
8 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
10 <title>event - Apache HTTP Server Version 2.5</title>
11 <link href="../style/css/manual.css" rel="stylesheet" media="all" type="text/css" title="Main stylesheet" />
12 <link href="../style/css/manual-loose-100pc.css" rel="alternate stylesheet" media="all" type="text/css" title="No Sidebar - Default font size" />
13 <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" />
14 <script src="../style/scripts/prettify.min.js" type="text/javascript">
17 <link href="../images/favicon.ico" rel="shortcut icon" /></head>
19 <div id="page-header">
20 <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>
21 <p class="apache">Apache HTTP Server Version 2.5</p>
22 <img alt="" src="../images/feather.png" /></div>
23 <div class="up"><a href="./"><img title="<-" alt="<-" src="../images/left.gif" /></a></div>
25 <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>
26 <div id="page-content">
27 <div id="preamble"><h1>Apache MPM event</h1>
29 <p><span>Available Languages: </span><a href="../en/mod/event.html" title="English"> en </a> |
30 <a href="../fr/mod/event.html" hreflang="fr" rel="alternate" title="Français"> fr </a></p>
32 <table class="module"><tr><th><a href="module-dict.html#Description">Description:</a></th><td>A variant of the <code class="module"><a href="../mod/worker.html">worker</a></code> MPM with the goal
33 of consuming threads only for connections with active processing</td></tr>
34 <tr><th><a href="module-dict.html#Status">Status:</a></th><td>MPM</td></tr>
35 <tr><th><a href="module-dict.html#ModuleIdentifier">Module Identifier:</a></th><td>mpm_event_module</td></tr>
36 <tr><th><a href="module-dict.html#SourceFile">Source File:</a></th><td>event.c</td></tr></table>
39 <p>The <code class="module"><a href="../mod/event.html">event</a></code> Multi-Processing Module (MPM) is,
40 as its name implies, an asynchronous, event-based implementation
41 designed to allow more requests to be served simultaneously by
42 passing off some processing work to the listeners threads, freeing up
43 the worker threads to serve new requests.</p>
45 <p>To use the <code class="module"><a href="../mod/event.html">event</a></code> MPM, add
46 <code>--with-mpm=event</code> to the <code class="program"><a href="../programs/configure.html">configure</a></code>
47 script's arguments when building the <code class="program"><a href="../programs/httpd.html">httpd</a></code>.</p>
50 <div id="quickview"><h3>Topics</h3>
52 <li><img alt="" src="../images/down.gif" /> <a href="#event-worker-relationship">Relationship with the Worker MPM</a></li>
53 <li><img alt="" src="../images/down.gif" /> <a href="#how-it-works">How it Works</a></li>
54 <li><img alt="" src="../images/down.gif" /> <a href="#requirements">Requirements</a></li>
55 </ul><h3 class="directives">Directives</h3>
57 <li><img alt="" src="../images/down.gif" /> <a href="#asyncrequestworkerfactor">AsyncRequestWorkerFactor</a></li>
58 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#coredumpdirectory">CoreDumpDirectory</a></li>
59 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#enableexceptionhook">EnableExceptionHook</a></li>
60 <li><img alt="" src="../images/right.gif" /> <a href="mod_unixd.html#group">Group</a></li>
61 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#listen">Listen</a></li>
62 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#listenbacklog">ListenBacklog</a></li>
63 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#maxconnectionsperchild">MaxConnectionsPerChild</a></li>
64 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#maxmemfree">MaxMemFree</a></li>
65 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#maxrequestworkers">MaxRequestWorkers</a></li>
66 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#maxsparethreads">MaxSpareThreads</a></li>
67 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#minsparethreads">MinSpareThreads</a></li>
68 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#pidfile">PidFile</a></li>
69 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#scoreboardfile">ScoreBoardFile</a></li>
70 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#sendbuffersize">SendBufferSize</a></li>
71 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#serverlimit">ServerLimit</a></li>
72 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#startservers">StartServers</a></li>
73 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#threadlimit">ThreadLimit</a></li>
74 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#threadsperchild">ThreadsPerChild</a></li>
75 <li><img alt="" src="../images/right.gif" /> <a href="mpm_common.html#threadstacksize">ThreadStackSize</a></li>
76 <li><img alt="" src="../images/right.gif" /> <a href="mod_unixd.html#user">User</a></li>
78 <h3>Bugfix checklist</h3><ul class="seealso"><li><a href="https://www.apache.org/dist/httpd/CHANGES_2.4">httpd changelog</a></li><li><a href="https://bz.apache.org/bugzilla/buglist.cgi?bug_status=__open__&list_id=144532&product=Apache%20httpd-2&query_format=specific&order=changeddate%20DESC%2Cpriority%2Cbug_severity&component=mpm_event">Known issues</a></li><li><a href="https://bz.apache.org/bugzilla/enter_bug.cgi?product=Apache%20httpd-2&component=mpm_event">Report a bug</a></li></ul><h3>See also</h3>
80 <li><a href="worker.html">The worker MPM</a></li>
81 <li><a href="#comments_section">Comments</a></li></ul></div>
82 <div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
84 <h2><a name="event-worker-relationship" id="event-worker-relationship">Relationship with the Worker MPM</a></h2>
85 <p><code class="module"><a href="../mod/event.html">event</a></code> is based on the <code class="module"><a href="../mod/worker.html">worker</a></code> MPM, which implements a hybrid
86 multi-process multi-threaded server. A single control process (the parent) is responsible for launching
87 child processes. Each child process creates a fixed number of server
88 threads as specified in the <code class="directive"><a href="../mod/mpm_common.html#threadsperchild">ThreadsPerChild</a></code> directive, as well
89 as a listener thread which listens for connections and passes them to a worker thread for processing when they arrive.</p>
91 <p>Run-time configuration directives are identical to those provided by <code class="module"><a href="../mod/worker.html">worker</a></code>, with the only addition
92 of the <code class="directive">AsyncRequestWorkerFactor</code>.</p>
94 </div><div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
96 <h2><a name="how-it-works" id="how-it-works">How it Works</a></h2>
97 <p>This original goal of this MPM was to fix the 'keep alive problem' in HTTP. After a client
98 completes the first request, it can keep the connection
99 open, sending further requests using the same socket and saving
100 significant overhead in creating TCP connections. However,
101 Apache HTTP Server traditionally keeps an entire child
102 process/thread waiting for data from the client, which brings its own disadvantages.
103 To solve this problem, this MPM uses a dedicated listener thread for each process
104 along with a pool of worker threads, sharing queues specific for those
105 requests in keep-alive mode (or, more simply, "readable"), those in write-
106 completion mode, and those in the process of shutting down ("closing").
107 An event loop, triggered on the status of the socket's availability,
108 adjusts these queues and pushes work to the worker pool.
111 <p>The total amount of connections that a single process/threads block can handle is regulated
112 by the <code class="directive">AsyncRequestWorkerFactor</code> directive.</p>
114 <h3><a name="async-connections" id="async-connections">Async connections</a></h3>
115 <p>Async connections would need a fixed dedicated worker thread with the previous MPMs but not with event.
116 The status page of <code class="module"><a href="../mod/mod_status.html">mod_status</a></code> shows new columns under the Async connections section:</p>
119 <dd>While sending the response to the client, it might happen that the TCP write buffer fills up because the connection is too slow. Usually in this case a <code>write()</code> to the socket returns <code>EWOULDBLOCK</code> or <code>EAGAIN</code>, to become writable again after an idle time. The worker holding the socket might be able to offload the waiting task to the listener thread, that in turn will re-assign it to the first idle worker thread available once an event will be raised for the socket (for example, "the socket is now writable"). Please check the Limitations section for more information.
123 <dd>Keep Alive handling is the most basic improvement from the worker MPM.
124 Once a worker thread finishes to flush the response to the client, it can offload the
125 socket handling to the listener thread, that in turns will wait for any event from the
126 OS, like "the socket is readable". If any new request comes from the client, then the
127 listener will forward it to the first worker thread available. Conversely, if the
128 <code class="directive"><a href="../mod/core.html#keepalivetimeout">KeepAliveTimeout</a></code> occurs then the socket will be
129 closed by the listener. In this way the worker threads are not responsible for idle
130 sockets and they can be re-used to serve other requests.</dd>
133 <dd>Sometimes the MPM needs to perform a lingering close, namely sending back an early error to the client while it is still transmitting data to httpd. Sending the response and then closing the connection immediately is not the correct thing to do since the client (still trying to send the rest of the request) would get a connection reset and could not read the httpd's response. So in such cases, httpd tries to read the rest of the request to allow the client to consume the response. The lingering close is time bounded but it can take relatively long time, so a worker thread can offload this work to the listener.</dd>
136 <p>These improvements are valid for both HTTP/HTTPS connections.</p>
140 <h3><a name="limitations" id="limitations">Limitations</a></h3>
141 <p>The improved connection handling may not work for certain connection
142 filters that have declared themselves as incompatible with event. In these
143 cases, this MPM will fall back to the behaviour of the
144 <code class="module"><a href="../mod/worker.html">worker</a></code> MPM and reserve one worker thread per connection.
145 All modules shipped with the server are compatible with the event MPM.</p>
147 <p>A similar restriction is currently present for requests involving an
148 output filter that needs to read and/or modify the whole response body.
149 If the connection to the client blocks while the filter is processing the
150 data, and the amount of data produced by the filter is too big to be
151 buffered in memory, the thread used for the request is not freed while
152 httpd waits until the pending data is sent to the client.<br />
153 To illustrate this point we can think about the following two situations:
154 serving a static asset (like a CSS file) versus serving content retrieved from
155 FCGI/CGI or a proxied server. The former is predictable, namely the event MPM
156 has full visibility on the end of the content and it can use events: the worker
157 thread serving the response content can flush the first bytes until <code>EWOULDBLOCK</code>
158 or <code>EAGAIN</code> is returned, delegating the rest to the listener. This one in turn
159 waits for an event on the socket, and delegates the work to flush the rest of the content
160 to the first idle worker thread. Meanwhile in the latter example (FCGI/CGI/proxied content)
161 the MPM can't predict the end of the response and a worker thread has to finish its work
162 before returning the control to the listener. The only alternative is to buffer the
163 response in memory, but it wouldn't be the safest option for the sake of the
164 server's stability and memory footprint.
169 <h3><a name="background" id="background">Background material</a></h3>
170 <p>The event model was made possible by the introduction of new APIs into the supported operating systems:</p>
172 <li>epoll (Linux) </li>
173 <li>kqueue (BSD) </li>
174 <li>event ports (Solaris) </li>
176 <p>Before these new APIs where made available, the traditional <code>select</code> and <code>poll</code> APIs had to be used.
177 Those APIs get slow if used to handle many connections or if the set of connections rate of change is high.
178 The new APIs allow to monitor much more connections and they perform way better when the set of connections to monitor changes frequently. So these APIs made it possible to write the event MPM, that scales much better with the typical HTTP pattern of many idle connections.</p>
180 <p>The MPM assumes that the underlying <code>apr_pollset</code>
181 implementation is reasonably threadsafe. This enables the MPM to
182 avoid excessive high level locking, or having to wake up the listener
183 thread in order to send it a keep-alive socket. This is currently
184 only compatible with KQueue and EPoll.</p>
188 </div><div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
189 <div class="section">
190 <h2><a name="requirements" id="requirements">Requirements</a></h2>
191 <p>This MPM depends on <a class="glossarylink" href="../glossary.html#apr" title="see glossary">APR</a>'s atomic
192 compare-and-swap operations for thread synchronization. If you are
193 compiling for an x86 target and you don't need to support 386s, or
194 you are compiling for a SPARC and you don't need to run on
195 pre-UltraSPARC chips, add
196 <code>--enable-nonportable-atomics=yes</code> to the
197 <code class="program"><a href="../programs/configure.html">configure</a></code> script's arguments. This will cause
198 APR to implement atomic operations using efficient opcodes not
199 available in older CPUs.</p>
201 <p>This MPM does not perform well on older platforms which lack good
202 threading, but the requirement for EPoll or KQueue makes this
207 <li>To use this MPM on FreeBSD, FreeBSD 5.3 or higher is recommended.
208 However, it is possible to run this MPM on FreeBSD 5.2.1, if you
209 use <code>libkse</code> (see <code>man libmap.conf</code>).</li>
211 <li>For NetBSD, at least version 2.0 is recommended.</li>
213 <li>For Linux, a 2.6 kernel is recommended. It is also necessary to
214 ensure that your version of <code>glibc</code> has been compiled
215 with support for EPoll.</li>
219 <div class="top"><a href="#page-header"><img alt="top" src="../images/up.gif" /></a></div>
220 <div class="directive-section"><h2><a name="AsyncRequestWorkerFactor" id="AsyncRequestWorkerFactor">AsyncRequestWorkerFactor</a> <a name="asyncrequestworkerfactor" id="asyncrequestworkerfactor">Directive</a></h2>
221 <table class="directive">
222 <tr><th><a href="directive-dict.html#Description">Description:</a></th><td>Limit concurrent connections per process</td></tr>
223 <tr><th><a href="directive-dict.html#Syntax">Syntax:</a></th><td><code>AsyncRequestWorkerFactor <var>factor</var></code></td></tr>
224 <tr><th><a href="directive-dict.html#Default">Default:</a></th><td><code>2</code></td></tr>
225 <tr><th><a href="directive-dict.html#Context">Context:</a></th><td>server config</td></tr>
226 <tr><th><a href="directive-dict.html#Status">Status:</a></th><td>MPM</td></tr>
227 <tr><th><a href="directive-dict.html#Module">Module:</a></th><td>event</td></tr>
228 <tr><th><a href="directive-dict.html#Compatibility">Compatibility:</a></th><td>Available in version 2.3.13 and later</td></tr>
230 <p>The event MPM handles some connections in an asynchronous way, where
231 request worker threads are only allocated for short periods of time as
232 needed, and other connections with one request worker thread reserved per
233 connection. This can lead to situations where all workers are tied up and
234 no worker thread is available to handle new work on established async
237 <p>To mitigate this problem, the event MPM does two things:</p>
239 <li>it limits the number of connections accepted per process, depending on the
240 number of idle request workers;</li>
241 <li>if all workers are busy, it will
242 close connections in keep-alive state even if the keep-alive timeout has
243 not expired. This allows the respective clients to reconnect to a
244 different process which may still have worker threads available.</li>
247 <p>This directive can be used to fine-tune the per-process connection
248 limit. A <strong>process</strong> will only accept new connections if the current number of
249 connections (not counting connections in the "closing" state) is lower
252 <p class="indent"><strong>
253 <code class="directive"><a href="../mod/mpm_common.html#threadsperchild">ThreadsPerChild</a></code> +
254 (<code class="directive">AsyncRequestWorkerFactor</code> *
255 <var>number of idle workers</var>)
258 <p>An estimation of the maximum concurrent connections across all the processes given
259 an average value of idle worker threads can be calculated with:
263 <p class="indent"><strong>
264 (<code class="directive"><a href="../mod/mpm_common.html#threadsperchild">ThreadsPerChild</a></code> +
265 (<code class="directive">AsyncRequestWorkerFactor</code> *
266 <var>number of idle workers</var>)) *
267 <code class="directive"><a href="../mod/mpm_common.html#serverlimit">ServerLimit</a></code>
270 <div class="note"><h3>Example</h3>
271 <pre class="prettyprint lang-config">ThreadsPerChild = 10
273 AsyncRequestWorkerFactor = 2
274 MaxRequestWorkers = 40
276 idle_workers = 4 (average for all the processes to keep it simple)
278 max_connections = (ThreadsPerChild + (AsyncRequestWorkerFactor * idle_workers)) * ServerLimit
279 = (10 + (2 * 4)) * 4 = 72</pre>
283 <p>When all the worker threads are idle, then absolute maximum numbers of concurrent
284 connections can be calculared in a simpler way:</p>
286 <p class="indent"><strong>
287 (<code class="directive">AsyncRequestWorkerFactor</code> + 1) *
288 <code class="directive"><a href="../mod/mpm_common.html#maxrequestworkers">MaxRequestWorkers</a></code>
292 <div class="note"><h3>Example</h3>
293 <pre class="prettyprint lang-config">ThreadsPerChild = 10
295 MaxRequestWorkers = 40
296 AsyncRequestWorkerFactor = 2</pre>
299 <p>If all the processes have all threads idle then: </p>
301 <pre class="prettyprint lang-config">idle_workers = 10</pre>
304 <p>We can calculate the absolute maximum numbers of concurrent connections in two ways:</p>
306 <pre class="prettyprint lang-config">max_connections = (ThreadsPerChild + (AsyncRequestWorkerFactor * idle_workers)) * ServerLimit
307 = (10 + (2 * 10)) * 4 = 120
309 max_connections = (AsyncRequestWorkerFactor + 1) * MaxRequestWorkers
310 = (2 + 1) * 40 = 120</pre>
314 <p>Tuning <code class="directive">AsyncRequestWorkerFactor</code> requires knowledge about the traffic handled by httpd in each specific use case, so changing the default value requires extensive testing and data gathering from <code class="module"><a href="../mod/mod_status.html">mod_status</a></code>.</p>
316 <p><code class="directive"><a href="../mod/mpm_common.html#maxrequestworkers">MaxRequestWorkers</a></code> was called
317 <code class="directive">MaxClients</code> prior to version 2.3.13. The above value
318 shows that the old name did not accurately describe its meaning for the event MPM.</p>
320 <p><code class="directive">AsyncRequestWorkerFactor</code> can take non-integer
321 arguments, e.g "1.5".</p>
326 <div class="bottomlang">
327 <p><span>Available Languages: </span><a href="../en/mod/event.html" title="English"> en </a> |
328 <a href="../fr/mod/event.html" hreflang="fr" rel="alternate" title="Français"> fr </a></p>
329 </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>
330 <script type="text/javascript"><!--//--><![CDATA[//><!--
331 var comments_shortname = 'httpd';
332 var comments_identifier = 'http://httpd.apache.org/docs/trunk/mod/event.html';
334 if (w.location.hostname.toLowerCase() == "httpd.apache.org") {
335 d.write('<div id="comments_thread"><\/div>');
336 var s = d.createElement('script');
337 s.type = 'text/javascript';
339 s.src = 'https://comments.apache.org/show_comments.lua?site=' + comments_shortname + '&page=' + comments_identifier;
340 (d.getElementsByTagName('head')[0] || d.getElementsByTagName('body')[0]).appendChild(s);
343 d.write('<div id="comments_thread">Comments are disabled for this page at the moment.<\/div>');
345 })(window, document);
346 //--><!]]></script></div><div id="footer">
347 <p class="apache">Copyright 2016 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>
348 <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"><!--//--><
projects / apache / blob