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 <!-- $LastChangedRevision$ -->
7 Copyright 2002-2005 The Apache Software Foundation or its licensors, as
10 Licensed under the Apache License, Version 2.0 (the "License");
11 you may not use this file except in compliance with the License.
12 You may obtain a copy of the License at
14 http://www.apache.org/licenses/LICENSE-2.0
16 Unless required by applicable law or agreed to in writing, software
17 distributed under the License is distributed on an "AS IS" BASIS,
18 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
19 See the License for the specific language governing permissions and
20 limitations under the License.
23 <manualpage metafile="stopping.xml.meta">
25 <title>Stopping and Restarting</title>
28 <p>This document covers stopping and restarting Apache on
29 Unix-like systems. Windows NT, 2000 and XP users should see
30 <a href="platform/windows.html#winsvc">Running Apache as a
31 Service</a> and Windows 9x and ME users should see <a
32 href="platform/windows.html#wincons">Running Apache as a
33 Console Application</a> for information on how to control
34 Apache on those platforms.</p>
37 <seealso><program>httpd</program></seealso>
38 <seealso><program>apachectl</program></seealso>
40 <section id="introduction"><title>Introduction</title>
42 <p>In order to stop or restart Apache, you must send a signal to
43 the running <program>httpd</program> processes. There are two ways to
44 send the signals. First, you can use the unix <code>kill</code>
45 command to directly send signals to the processes. You will
46 notice many <program>httpd</program> executables running on your system,
47 but you should not send signals to any of them except the parent,
48 whose pid is in the <directive
49 module="mpm_common">PidFile</directive>. That is to say you
50 shouldn't ever need to send signals to any process except the
51 parent. There are three signals that you can send the parent:
52 <code><a href="#term">TERM</a></code>,
53 <code><a href="#hup">HUP</a></code>, and
54 <code><a href="#graceful">USR1</a></code>, which
55 will be described in a moment.</p>
57 <p>To send a signal to the parent you should issue a command
60 <example>kill -TERM `cat /usr/local/apache2/logs/httpd.pid`</example>
62 <p>The second method of signaling the <program>httpd</program> processes
63 is to use the <code>-k</code> command line options: <code>stop</code>,
64 <code>restart</code>, and <code>graceful</code>,
65 as described below. These are arguments to the <program>
66 httpd</program> binary, but we recommend that
67 you send them using the <program>apachectl</program> control script, which
68 will pass them through to <program>httpd</program>.</p>
70 <p>After you have signaled <program>httpd</program>, you can read about
71 its progress by issuing:</p>
73 <example>tail -f /usr/local/apache2/logs/error_log</example>
75 <p>Modify those examples to match your <directive
76 module="core">ServerRoot</directive> and <directive
77 module="mpm_common">PidFile</directive> settings.</p>
80 <section id="term"><title>Stop Now</title>
82 <dl><dt>Signal: TERM</dt>
83 <dd><code>apachectl -k stop</code></dd>
86 <p>Sending the <code>TERM</code> or <code>stop</code> signal to
87 the parent causes it to immediately attempt to kill off all of its
88 children. It may take it several seconds to complete killing off
89 its children. Then the parent itself exits. Any requests in
90 progress are terminated, and no further requests are served.</p>
93 <section id="graceful"><title>Graceful Restart</title>
95 <dl><dt>Signal: USR1</dt>
96 <dd><code>apachectl -k graceful</code></dd>
99 <p>The <code>USR1</code> or <code>graceful</code> signal causes
100 the parent process to <em>advise</em> the children to exit after
101 their current request (or to exit immediately if they're not
102 serving anything). The parent re-reads its configuration files and
103 re-opens its log files. As each child dies off the parent replaces
104 it with a child from the new <em>generation</em> of the
105 configuration, which begins serving new requests immediately.</p>
107 <note>On certain platforms that do not allow <code>USR1</code> to
108 be used for a graceful restart, an alternative signal may be used (such
109 as <code>WINCH</code>). The command <code>apachectl graceful</code>
110 will send the right signal for your platform.</note>
112 <p>This code is designed to always respect the process control
113 directive of the MPMs, so the number of processes and threads
114 available to serve clients will be maintained at the appropriate
115 values throughout the restart process. Furthermore, it respects
116 <directive module="mpm_common">StartServers</directive> in the
117 following manner: if after one second at least <directive
118 module="mpm_common">StartServers</directive> new children have not
119 been created, then create enough to pick up the slack. Hence the
120 code tries to maintain both the number of children appropriate for
121 the current load on the server, and respect your wishes with the
122 <directive module="mpm_common">StartServers</directive>
125 <p>Users of <module>mod_status</module>
126 will notice that the server statistics are <strong>not</strong>
127 set to zero when a <code>USR1</code> is sent. The code was
128 written to both minimize the time in which the server is unable
129 to serve new requests (they will be queued up by the operating
130 system, so they're not lost in any event) and to respect your
131 tuning parameters. In order to do this it has to keep the
132 <em>scoreboard</em> used to keep track of all children across
135 <p>The status module will also use a <code>G</code> to indicate
136 those children which are still serving requests started before
137 the graceful restart was given.</p>
139 <p>At present there is no way for a log rotation script using
140 <code>USR1</code> to know for certain that all children writing
141 the pre-restart log have finished. We suggest that you use a
142 suitable delay after sending the <code>USR1</code> signal
143 before you do anything with the old log. For example if most of
144 your hits take less than 10 minutes to complete for users on
145 low bandwidth links then you could wait 15 minutes before doing
146 anything with the old log.</p>
148 <note>If your configuration file has errors
149 in it when you issue a restart then your parent will not
150 restart, it will exit with an error. In the case of graceful
151 restarts it will also leave children running when it exits.
152 (These are the children which are "gracefully exiting" by
153 handling their last request.) This will cause problems if you
154 attempt to restart the server -- it will not be able to bind to
155 its listening ports. Before doing a restart, you can check the
156 syntax of the configuration files with the <code>-t</code>
157 command line argument (see <program>httpd</program>). This still will not
158 guarantee that the server will restart correctly. To check the
159 semantics of the configuration files as well as the syntax, you
160 can try starting <program>httpd</program> as a non-root user. If there
161 are no errors it will attempt to open its sockets and logs and fail
162 because it's not root (or because the currently running
163 <program>httpd</program> already has those ports bound). If it fails
164 for any other reason then it's probably a config file error and the error
165 should be fixed before issuing the graceful restart.</note>
168 <section id="hup"><title>Restart Now</title>
170 <dl><dt>Signal: HUP</dt>
171 <dd><code>apachectl -k restart</code></dd>
174 <p>Sending the <code>HUP</code> or <code>restart</code> signal to
175 the parent causes it to kill off its children like in
176 <code>TERM</code>, but the parent doesn't exit. It re-reads its
177 configuration files, and re-opens any log files. Then it spawns a
178 new set of children and continues serving hits.</p>
180 <p>Users of <module>mod_status</module>
181 will notice that the server statistics are set to zero when a
182 <code>HUP</code> is sent.</p>
184 <note>If your configuration file has errors in it when you issue a
185 restart then your parent will not restart, it will exit with an
186 error. See above for a method of avoiding this.</note>
189 <section id="race"><title>Appendix: signals and race conditions</title>
191 <p>Prior to Apache 1.2b9 there were several <em>race
192 conditions</em> involving the restart and die signals (a simply put,
193 a race condition is a time-sensitive problem - if something happens
194 at just the wrong time or things happen in the wrong order,
195 undesired behaviour will result. If the same thing happens at the right
196 time, all will be well). For those architectures that have the "right"
197 feature set we have eliminated as many as we can. But it should
198 be noted that race conditions do still exist on certain
201 <p>Architectures that use an on-disk <directive
202 module="mpm_common">ScoreBoardFile</directive> can potentially have
203 their scoreboards corrupted. This can result in the "bind:
204 Address already in use" (after <code>HUP</code>) or "long lost
205 child came home!" (after <code>USR1</code>). The former is a fatal
206 error, while the latter just causes the server to lose a
207 scoreboard slot. So it may be advisable to use graceful
208 restarts, with an occasional hard restart. These problems are very
209 difficult to work around, but fortunately most architectures do
210 not require a scoreboard file. See the <directive
211 module="mpm_common">ScoreBoardFile</directive> documentation for
212 architecture which uses it.</p>
214 <p>All architectures have a small race condition in each child
215 involving the second and subsequent requests on a persistent
216 HTTP connection (KeepAlive). It may exit after reading the
217 request line but before reading any of the request headers.
218 There is a fix that was discovered too late to make 1.2. In
219 theory this isn't an issue because the KeepAlive client has to
220 expect these events because of network latencies and server
221 timeouts. In practice it doesn't seem to affect anything either
222 -- in a test case the server was restarted twenty times per
223 second and clients successfully browsed the site without
224 getting broken images or empty documents. </p>