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 2002-2004 The Apache Software Foundation
9 Licensed under the Apache License, Version 2.0 (the "License");
10 you may not use this file except in compliance with the License.
11 You may obtain a copy of the License at
13 http://www.apache.org/licenses/LICENSE-2.0
15 Unless required by applicable law or agreed to in writing, software
16 distributed under the License is distributed on an "AS IS" BASIS,
17 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
18 See the License for the specific language governing permissions and
19 limitations under the License.
22 <manualpage metafile="stopping.xml.meta">
24 <title>Stopping and Restarting</title>
27 <p>This document covers stopping and restarting Apache on
28 Unix-like systems. Windows NT, 2000 and XP users should see
29 <a href="platform/windows.html#winsvc">Running Apache as a
30 Service</a> and Windows 9x and ME users should see <a
31 href="platform/windows.html#wincons">Running Apache as a
32 Console Application</a> for information on how to control
33 Apache on those platforms.</p>
36 <seealso><program>httpd</program></seealso>
37 <seealso><program>apachectl</program></seealso>
39 <section id="introduction"><title>Introduction</title>
41 <p>In order to stop or restart Apache, you must send a signal to
42 the running <program>httpd</program> processes. There are two ways to
43 send the signals. First, you can use the unix <code>kill</code>
44 command to directly send signals to the processes. You will
45 notice many <program>httpd</program> executables running on your system,
46 but you should not send signals to any of them except the parent,
47 whose pid is in the <directive
48 module="mpm_common">PidFile</directive>. That is to say you
49 shouldn't ever need to send signals to any process except the
50 parent. There are three signals that you can send the parent:
51 <code><a href="#term">TERM</a></code>,
52 <code><a href="#hup">HUP</a></code>, and
53 <code><a href="#graceful">USR1</a></code>, which
54 will be described in a moment.</p>
56 <p>To send a signal to the parent you should issue a command
59 <example>kill -TERM `cat /usr/local/apache2/logs/httpd.pid`</example>
61 <p>The second method of signaling the <program>httpd</program> processes
62 is to use the <code>-k</code> command line options: <code>stop</code>,
63 <code>restart</code>, and <code>graceful</code>,
64 as described below. These are arguments to the <program>
65 httpd</program> binary, but we recommend that
66 you send them using the <program>apachectl</program> control script, which
67 will pass them through to <program>httpd</program>.</p>
69 <p>After you have signaled <program>httpd</program>, you can read about
70 its progress by issuing:</p>
72 <example>tail -f /usr/local/apache2/logs/error_log</example>
74 <p>Modify those examples to match your <directive
75 module="core">ServerRoot</directive> and <directive
76 module="mpm_common">PidFile</directive> settings.</p>
79 <section id="term"><title>Stop Now</title>
81 <dl><dt>Signal: TERM</dt>
82 <dd><code>apachectl -k stop</code></dd>
85 <p>Sending the <code>TERM</code> or <code>stop</code> signal to
86 the parent causes it to immediately attempt to kill off all of its
87 children. It may take it several seconds to complete killing off
88 its children. Then the parent itself exits. Any requests in
89 progress are terminated, and no further requests are served.</p>
92 <section id="graceful"><title>Graceful Restart</title>
94 <dl><dt>Signal: USR1</dt>
95 <dd><code>apachectl -k graceful</code></dd>
98 <p>The <code>USR1</code> or <code>graceful</code> signal causes
99 the parent process to <em>advise</em> the children to exit after
100 their current request (or to exit immediately if they're not
101 serving anything). The parent re-reads its configuration files and
102 re-opens its log files. As each child dies off the parent replaces
103 it with a child from the new <em>generation</em> of the
104 configuration, which begins serving new requests immediately.</p>
106 <note>On certain platforms that do not allow <code>USR1</code> to
107 be used for a graceful restart, an alternative signal may be used (such
108 as <code>WINCH</code>). The command <code>apachectl graceful</code>
109 will send the right signal for your platform.</note>
111 <p>This code is designed to always respect the process control
112 directive of the MPMs, so the number of processes and threads
113 available to serve clients will be maintained at the appropriate
114 values throughout the restart process. Furthermore, it respects
115 <directive module="mpm_common">StartServers</directive> in the
116 following manner: if after one second at least <directive
117 module="mpm_common">StartServers</directive> new children have not
118 been created, then create enough to pick up the slack. Hence the
119 code tries to maintain both the number of children appropriate for
120 the current load on the server, and respect your wishes with the
121 <directive>StartServers</directive> parameter.</p>
123 <p>Users of the <module>mod_status</module>
124 will notice that the server statistics are <strong>not</strong>
125 set to zero when a <code>USR1</code> is sent. The code was
126 written to both minimize the time in which the server is unable
127 to serve new requests (they will be queued up by the operating
128 system, so they're not lost in any event) and to respect your
129 tuning parameters. In order to do this it has to keep the
130 <em>scoreboard</em> used to keep track of all children across
133 <p>The status module will also use a <code>G</code> to indicate
134 those children which are still serving requests started before
135 the graceful restart was given.</p>
137 <p>At present there is no way for a log rotation script using
138 <code>USR1</code> to know for certain that all children writing
139 the pre-restart log have finished. We suggest that you use a
140 suitable delay after sending the <code>USR1</code> signal
141 before you do anything with the old log. For example if most of
142 your hits take less than 10 minutes to complete for users on
143 low bandwidth links then you could wait 15 minutes before doing
144 anything with the old log.</p>
146 <note>If your configuration file has errors
147 in it when you issue a restart then your parent will not
148 restart, it will exit with an error. In the case of graceful
149 restarts it will also leave children running when it exits.
150 (These are the children which are "gracefully exiting" by
151 handling their last request.) This will cause problems if you
152 attempt to restart the server -- it will not be able to bind to
153 its listening ports. Before doing a restart, you can check the
154 syntax of the configuration files with the <code>-t</code>
155 command line argument (see <program>httpd</program>). This still will not
156 guarantee that the server will restart correctly. To check the
157 semantics of the configuration files as well as the syntax, you
158 can try starting <program>httpd</program> as a non-root user. If there
159 are no errors it will attempt to open its sockets and logs and fail
160 because it's not root (or because the currently running
161 <program>httpd</program> already has those ports bound). If it fails
162 for any other reason then it's probably a config file error and the error
163 should be fixed before issuing the graceful restart.</note>
166 <section id="hup"><title>Restart Now</title>
168 <dl><dt>Signal: HUP</dt>
169 <dd><code>apachectl -k restart</code></dd>
172 <p>Sending the <code>HUP</code> or <code>restart</code> signal to
173 the parent causes it to kill off its children like in
174 <code>TERM</code>, but the parent doesn't exit. It re-reads its
175 configuration files, and re-opens any log files. Then it spawns a
176 new set of children and continues serving hits.</p>
178 <p>Users of <module>mod_status</module>
179 will notice that the server statistics are set to zero when a
180 <code>HUP</code> is sent.</p>
182 <note>If your configuration file has errors in it when you issue a
183 restart then your parent will not restart, it will exit with an
184 error. See above for a method of avoiding this.</note>
187 <section id="race"><title>Appendix: signals and race conditions</title>
189 <p>Prior to Apache 1.2b9 there were several <em>race
190 conditions</em> involving the restart and die signals (a simple
191 description of race condition is: a time-sensitive problem, as
192 in if something happens at just the wrong time it won't behave
193 as expected). For those architectures that have the "right"
194 feature set we have eliminated as many as we can. But it should
195 be noted that there still do exist race conditions on certain
198 <p>Architectures that use an on disk <directive
199 module="mpm_common">ScoreBoardFile</directive> have the potential
200 to corrupt their scoreboards. This can result in the "bind:
201 Address already in use" (after <code>HUP</code>) or "long lost
202 child came home!" (after <code>USR1</code>). The former is a fatal
203 error, while the latter just causes the server to lose a
204 scoreboard slot. So it might be advisable to use graceful
205 restarts, with an occasional hard restart. These problems are very
206 difficult to work around, but fortunately most architectures do
207 not require a scoreboard file. See the <directive
208 module="mpm_common">ScoreBoardFile</directive> documentation for a
209 architecture uses it.</p>
211 <p>All architectures have a small race condition in each child
212 involving the second and subsequent requests on a persistent
213 HTTP connection (KeepAlive). It may exit after reading the
214 request line but before reading any of the request headers.
215 There is a fix that was discovered too late to make 1.2. In
216 theory this isn't an issue because the KeepAlive client has to
217 expect these events because of network latencies and server
218 timeouts. In practice it doesn't seem to affect anything either
219 -- in a test case the server was restarted twenty times per
220 second and clients successfully browsed the site without
221 getting broken images or empty documents. </p>