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"?>
4 <manualpage metafile="stopping.xml.meta">
6 <title>Stopping and Restarting</title>
9 <p>This document covers stopping and restarting Apache on
10 Unix-like systems. Windows NT, 2000 and XP users should see
11 <a href="platform/windows.html#winsvc">Running Apache as a
12 Service</a> and Windows 9x and ME users should see <a
13 href="platform/windows.html#wincons">Running Apache as a
14 Console Application</a> for information on how to control
15 Apache on those platforms.</p>
18 <seealso><a href="programs/httpd.html">httpd</a></seealso>
19 <seealso><a href="programs/apachectl.html">apachectl</a></seealso>
21 <section id="introduction"><title>Introduction</title>
23 <p>In order to stop or restart Apache, you must send a signal to
24 the running <code>httpd</code> processes. There are two ways to
25 send the signals. First, you can use the unix <code>kill</code>
26 command to directly send signals to the processes. You will
27 notice many <code>httpd</code> executables running on your system,
28 but you should not send signals to any of them except the parent,
29 whose pid is in the <directive
30 module="mpm_common">PidFile</directive>. That is to say you
31 shouldn't ever need to send signals to any process except the
32 parent. There are three signals that you can send the parent:
33 <code><a href="#term">TERM</a></code>,
34 <code><a href="#hup">HUP</a></code>, and
35 <code><a href="#graceful">USR1</a></code>, which
36 will be described in a moment.</p>
38 <p>To send a signal to the parent you should issue a command
41 <example>kill -TERM `cat /usr/local/apache2/logs/httpd.pid`</example>
43 <p>The second method of signaling the <code>httpd</code> processes
44 is to use the <code>-k</code> command line options: <code>stop</code>,
45 <code>restart</code>, and <code>graceful</code>,
46 as described below. These are arguments to the <a
47 href="programs/httpd.html">httpd</a> binary, but we recommend that
48 you send them using the <a
49 href="programs/apachectl.html">apachectl</a> control script, which
50 will pass them through to <code>httpd</code>.</p>
52 <p>After you have signaled <code>httpd</code>, you can read about
53 its progress by issuing:</p>
55 <example>tail -f /usr/local/apache2/logs/error_log</example>
57 <p>Modify those examples to match your <directive
58 module="core">ServerRoot</directive> and <directive
59 module="mpm_common">PidFile</directive> settings.</p>
62 <section id="term"><title>Stop Now</title>
64 <dl><dt>Signal: TERM</dt>
65 <dd><code>apachectl -k stop</code></dd>
68 <p>Sending the <code>TERM</code> or <code>stop</code> signal to
69 the parent causes it to immediately attempt to kill off all of its
70 children. It may take it several seconds to complete killing off
71 its children. Then the parent itself exits. Any requests in
72 progress are terminated, and no further requests are served.</p>
75 <section id="graceful"><title>Graceful Restart</title>
77 <dl><dt>Signal: USR1</dt>
78 <dd><code>apachectl -k graceful</code></dd>
81 <p>The <code>USR1</code> or <code>graceful</code> signal causes
82 the parent process to <em>advise</em> the children to exit after
83 their current request (or to exit immediately if they're not
84 serving anything). The parent re-reads its configuration files and
85 re-opens its log files. As each child dies off the parent replaces
86 it with a child from the new <em>generation</em> of the
87 configuration, which begins serving new requests immediately.</p>
89 <note>On certain platforms that do not allow <code>USR1</code> to
90 be used for a graceful restart, an alternative signal may be used (such
91 as <code>WINCH</code>). The command <code>apachectl graceful</code>
92 will send the right signal for your platform.</note>
94 <p>This code is designed to always respect the process control
95 directive of the MPMs, so the number of processes and threads
96 available to serve clients will be maintained at the appropriate
97 values throughout the restart process. Furthermore, it respects
98 <directive module="mpm_common">StartServers</directive> in the
99 following manner: if after one second at least <directive
100 module="mpm_common">StartServers</directive> new children have not
101 been created, then create enough to pick up the slack. Hence the
102 code tries to maintain both the number of children appropriate for
103 the current load on the server, and respect your wishes with the
104 <directive>StartServers</directive> parameter.</p>
106 <p>Users of the <module>mod_status</module>
107 will notice that the server statistics are <strong>not</strong>
108 set to zero when a <code>USR1</code> is sent. The code was
109 written to both minimize the time in which the server is unable
110 to serve new requests (they will be queued up by the operating
111 system, so they're not lost in any event) and to respect your
112 tuning parameters. In order to do this it has to keep the
113 <em>scoreboard</em> used to keep track of all children across
116 <p>The status module will also use a <code>G</code> to indicate
117 those children which are still serving requests started before
118 the graceful restart was given.</p>
120 <p>At present there is no way for a log rotation script using
121 <code>USR1</code> to know for certain that all children writing
122 the pre-restart log have finished. We suggest that you use a
123 suitable delay after sending the <code>USR1</code> signal
124 before you do anything with the old log. For example if most of
125 your hits take less than 10 minutes to complete for users on
126 low bandwidth links then you could wait 15 minutes before doing
127 anything with the old log.</p>
129 <note>If your configuration file has errors
130 in it when you issue a restart then your parent will not
131 restart, it will exit with an error. In the case of graceful
132 restarts it will also leave children running when it exits.
133 (These are the children which are "gracefully exiting" by
134 handling their last request.) This will cause problems if you
135 attempt to restart the server -- it will not be able to bind to
136 its listening ports. Before doing a restart, you can check the
137 syntax of the configuration files with the <code>-t</code>
138 command line argument (see <a
139 href="programs/httpd.html">httpd</a>). This still will not
140 guarantee that the server will restart correctly. To check the
141 semantics of the configuration files as well as the syntax, you
142 can try starting <code>httpd</code> as a non-root user. If there are no
143 errors it will attempt to open its sockets and logs and fail
144 because it's not root (or because the currently running <code>httpd</code>
145 already has those ports bound). If it fails for any other
146 reason then it's probably a config file error and the error
147 should be fixed before issuing the graceful restart.</note>
150 <section id="hup"><title>Restart Now</title>
152 <dl><dt>Signal: HUP</dt>
153 <dd><code>apachectl -k restart</code></dd>
156 <p>Sending the <code>HUP</code> or <code>restart</code> signal to
157 the parent causes it to kill off its children like in
158 <code>TERM</code>, but the parent doesn't exit. It re-reads its
159 configuration files, and re-opens any log files. Then it spawns a
160 new set of children and continues serving hits.</p>
162 <p>Users of <module>mod_status</module>
163 will notice that the server statistics are set to zero when a
164 <code>HUP</code> is sent.</p>
166 <note>If your configuration file has errors in it when you issue a
167 restart then your parent will not restart, it will exit with an
168 error. See above for a method of avoiding this.</note>
171 <section id="race"><title>Appendix: signals and race conditions</title>
173 <p>Prior to Apache 1.2b9 there were several <em>race
174 conditions</em> involving the restart and die signals (a simple
175 description of race condition is: a time-sensitive problem, as
176 in if something happens at just the wrong time it won't behave
177 as expected). For those architectures that have the "right"
178 feature set we have eliminated as many as we can. But it should
179 be noted that there still do exist race conditions on certain
182 <p>Architectures that use an on disk <directive
183 module="mpm_common">ScoreBoardFile</directive> have the potential
184 to corrupt their scoreboards. This can result in the "bind:
185 Address already in use" (after <code>HUP</code>) or "long lost
186 child came home!" (after <code>USR1</code>). The former is a fatal
187 error, while the latter just causes the server to lose a
188 scoreboard slot. So it might be advisable to use graceful
189 restarts, with an occasional hard restart. These problems are very
190 difficult to work around, but fortunately most architectures do
191 not require a scoreboard file. See the <directive
192 module="mpm_common">ScoreBoardFile</directive> documentation for a
193 architecture uses it.</p>
195 <p>All architectures have a small race condition in each child
196 involving the second and subsequent requests on a persistent
197 HTTP connection (KeepAlive). It may exit after reading the
198 request line but before reading any of the request headers.
199 There is a fix that was discovered too late to make 1.2. In
200 theory this isn't an issue because the KeepAlive client has to
201 expect these events because of network latencies and server
202 timeouts. In practice it doesn't seem to affect anything either
203 -- in a test case the server was restarted twenty times per
204 second and clients successfully browsed the site without
205 getting broken images or empty documents. </p>