1 <?xml version='1.0' encoding='UTF-8' ?>
2 <!DOCTYPE manualpage SYSTEM "./style/manualpage.dtd">
3 <?xml-stylesheet type="text/xsl" href="./style/manual.en.xsl"?>
5 <relativepath href="."/>
7 <title>Stopping and Restarting</title>
10 <p>This document covers stopping and restarting Apache on
11 Unix-like systems. Windows users should see <a
12 href="platform/windows.html#signal">Signaling Apache when
16 <seealso><a href="programs/httpd.html">httpd</a></seealso>
17 <seealso><a href="programs/apachectl.html">apachectl</a></seealso>
19 <section id="introduction"><title>Introduction</title>
21 <p>In order to stop or restart Apache, you must send a signal to
22 the running <code>httpd</code> processes. There are two ways to
23 send the signals. First, you can use the unix <code>kill</code>
24 command to directly send signals to the processes. You will
25 notice many <code>httpd</code> executables running on your system,
26 but you should not send signals to any of them except the parent,
27 whose pid is in the <directive
28 module="mpm_common">PidFile</directive>. That is to say you
29 shouldn't ever need to send signals to any process except the
30 parent. There are three signals that you can send the parent:
31 <code>TERM</code>, <code>HUP</code>, and <code>USR1</code>, which
32 will be described in a moment.</p>
34 <p>To send a signal to the parent you should issue a command
37 <example>kill -TERM `cat /usr/local/apache2/logs/httpd.pid`</example>
39 <p>The second method of signaling the <code>httpd</code> processes
40 is to use the <code>-k</code> command line options: <code>stop</code>,
41 <code>restart</code>, and <code>graceful</code>,
42 as described below. These are arguments to the <a
43 href="programs/httpd.html">httpd</a> binary, but we recommend that
44 you send them using the <a
45 href="programs/apachectl.html">apachectl</a> control script, which
46 will pass them through to <code>httpd</code>.</p>
48 <p>After you have signaled <code>httpd</code>, you can read about
49 its progress by issuing:</p>
51 <example>tail -f /usr/local/apache2/logs/error_log</example>
53 <p>Modify those examples to match your <directive
54 module="core">ServerRoot</directive> and <directive
55 module="mpm_common">PidFile</directive> settings.</p>
58 <section id="term"><title>Stop Now</title>
60 <dl><dt>Signal: TERM</dt>
61 <dd><code>apachectl -k stop</code></dd>
64 <p>Sending the <code>TERM</code> or <code>stop</code> signal to
65 the parent causes it to immediately attempt to kill off all of its
66 children. It may take it several seconds to complete killing off
67 its children. Then the parent itself exits. Any requests in
68 progress are terminated, and no further requests are served.</p>
71 <section id="graceful"><title>Graceful Restart</title>
73 <dl><dt>Signal: USR1</dt>
74 <dd><code>apachectl -k graceful</code></dd>
77 <p>The <code>USR1</code> or <code>graceful</code> signal causes
78 the parent process to <em>advise</em> the children to exit after
79 their current request (or to exit immediately if they're not
80 serving anything). The parent re-reads its configuration files and
81 re-opens its log files. As each child dies off the parent replaces
82 it with a child from the new <em>generation</em> of the
83 configuration, which begins serving new requests immediately.</p>
85 <note>On certain platforms that do not allow <code>USR1</code> to
86 be used for a graceful restart, an alternative signal may be used (such
87 as <code>WINCH</code>). The command <code>apachectl graceful</code>
88 will send the right signal for your platform.</note>
90 <p>This code is designed to always respect the process control
91 directive of the MPMs, so the number of processes and threads
92 available to serve clients will be maintained at the appropriate
93 values throughout the restart process. Furthermore, it respects
94 <directive module="mpm_common">StartServers</directive> in the
95 following manner: if after one second at least <directive
96 module="mpm_common">StartServers</directive> new children have not
97 been created, then create enough to pick up the slack. Hence the
98 code tries to maintain both the number of children appropriate for
99 the current load on the server, and respect your wishes with the
100 <directive>StartServers</directive> parameter.</p>
102 <p>Users of the <module>mod_status</module>
103 will notice that the server statistics are <strong>not</strong>
104 set to zero when a <code>USR1</code> is sent. The code was
105 written to both minimize the time in which the server is unable
106 to serve new requests (they will be queued up by the operating
107 system, so they're not lost in any event) and to respect your
108 tuning parameters. In order to do this it has to keep the
109 <em>scoreboard</em> used to keep track of all children across
112 <p>The status module will also use a <code>G</code> to indicate
113 those children which are still serving requests started before
114 the graceful restart was given.</p>
116 <p>At present there is no way for a log rotation script using
117 <code>USR1</code> to know for certain that all children writing
118 the pre-restart log have finished. We suggest that you use a
119 suitable delay after sending the <code>USR1</code> signal
120 before you do anything with the old log. For example if most of
121 your hits take less than 10 minutes to complete for users on
122 low bandwidth links then you could wait 15 minutes before doing
123 anything with the old log.</p>
125 <note>If your configuration file has errors
126 in it when you issue a restart then your parent will not
127 restart, it will exit with an error. In the case of graceful
128 restarts it will also leave children running when it exits.
129 (These are the children which are "gracefully exiting" by
130 handling their last request.) This will cause problems if you
131 attempt to restart the server -- it will not be able to bind to
132 its listening ports. Before doing a restart, you can check the
133 syntax of the configuration files with the <code>-t</code>
134 command line argument (see <a
135 href="programs/httpd.html">httpd</a>). This still will not
136 guarantee that the server will restart correctly. To check the
137 semantics of the configuration files as well as the syntax, you
138 can try starting <code>httpd</code> as a non-root user. If there are no
139 errors it will attempt to open its sockets and logs and fail
140 because it's not root (or because the currently running <code>httpd</code>
141 already has those ports bound). If it fails for any other
142 reason then it's probably a config file error and the error
143 should be fixed before issuing the graceful restart.</note>
146 <section id="hup"><title>Restart Now</title>
148 <dl><dt>Signal: HUP</dt>
149 <dd><code>apachectl -k restart</code></dd>
152 <p>Sending the <code>HUP</code> or <code>restart</code> signal to
153 the parent causes it to kill off its children like in
154 <code>TERM</code>, but the parent doesn't exit. It re-reads its
155 configuration files, and re-opens any log files. Then it spawns a
156 new set of children and continues serving hits.</p>
158 <p>Users of <module>mod_status</module>
159 will notice that the server statistics are set to zero when a
160 <code>HUP</code> is sent.</p>
162 <note>If your configuration file has errors in it when you issue a
163 restart then your parent will not restart, it will exit with an
164 error. See above for a method of avoiding this.</note>
167 <section id="race"><title>Appendix: signals and race conditions</title>
169 <p>Prior to Apache 1.2b9 there were several <em>race
170 conditions</em> involving the restart and die signals (a simple
171 description of race condition is: a time-sensitive problem, as
172 in if something happens at just the wrong time it won't behave
173 as expected). For those architectures that have the "right"
174 feature set we have eliminated as many as we can. But it should
175 be noted that there still do exist race conditions on certain
178 <p>Architectures that use an on disk <directive
179 module="mpm_common">ScoreBoardFile</directive> have the potential
180 to corrupt their scoreboards. This can result in the "bind:
181 Address already in use" (after <code>HUP</code>) or "long lost
182 child came home!" (after <code>USR1</code>). The former is a fatal
183 error, while the latter just causes the server to lose a
184 scoreboard slot. So it might be advisable to use graceful
185 restarts, with an occasional hard restart. These problems are very
186 difficult to work around, but fortunately most architectures do
187 not require a scoreboard file. See the <directive
188 module="mpm_common">ScoreBoardFile</directive> documentation for a
189 architecture uses it.</p>
191 <p>All architectures have a small race condition in each child
192 involving the second and subsequent requests on a persistent
193 HTTP connection (KeepAlive). It may exit after reading the
194 request line but before reading any of the request headers.
195 There is a fix that was discovered too late to make 1.2. In
196 theory this isn't an issue because the KeepAlive client has to
197 expect these events because of network latencies and server
198 timeouts. In practice it doesn't seem to affect anything either
199 -- in a test case the server was restarted twenty times per
200 second and clients successfully browsed the site without
201 getting broken images or empty documents. </p>