chiark / gitweb /
update TODO
[elogind.git] / man / systemd.service.xml
index 7b6f12d069bd18ca5f26d711e720c024cf7233c4..258b059efa26b45f4a72ccfaf93d08642504c721 100644 (file)
@@ -9,16 +9,16 @@
   Copyright 2010 Lennart Poettering
 
   systemd is free software; you can redistribute it and/or modify it
-  under the terms of the GNU General Public License as published by
-  the Free Software Foundation; either version 2 of the License, or
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
   (at your option) any later version.
 
   systemd is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-  General Public License for more details.
+  Lesser General Public License for more details.
 
-  You should have received a copy of the GNU General Public License
+  You should have received a copy of the GNU Lesser General Public License
   along with systemd; If not, see <http://www.gnu.org/licenses/>.
 -->
 
                                 below) should be set to open access to
                                 the notification socket provided by
                                 systemd. If
-                                <varname>NotifyAccess=</varname> is not
-                                set, it will implicitly be set to
+                                <varname>NotifyAccess=</varname> is
+                                not set, it will be implicitly set to
                                 <option>main</option>.</para>
                                 </listitem>
                         </varlistentry>
                                 90s.</para></listitem>
                         </varlistentry>
 
+                        <varlistentry>
+                                <term><varname>WatchdogSec=</varname></term>
+                                <listitem><para>Configures the
+                                watchdog timeout for a service. This
+                                is activated when the start-up is
+                                completed. The service must call
+                                <citerefentry><refentrytitle>sd_notify</refentrytitle><manvolnum>3</manvolnum></citerefentry>
+                                regularly with "WATCHDOG=1". If the
+                                time between two such calls is larger
+                                than the configured time then the
+                                service is placed in a failure
+                                state. By setting
+                                <varname>Restart=</varname>
+                                to <option>on-failure</option> or
+                                <option>always</option> the service
+                                will be automatically restarted. The
+                                time configured here will be passed to
+                                the executed service process in the
+                                <varname>WATCHDOG_USEC=</varname>
+                                environment variable. If
+                                this option is used
+                                <varname>NotifyAccess=</varname> (see
+                                below) should be set to open access to
+                                the notification socket provided by
+                                systemd. If
+                                <varname>NotifyAccess=</varname> is not
+                                set, it will be implicitly set to
+                                <option>main</option>. Defaults to 0,
+                                which disables this
+                                feature.</para></listitem>
+                        </varlistentry>
+
                         <varlistentry>
                                 <term><varname>Restart=</varname></term>
                                 <listitem><para>Configures whether the
                                 0. If set to
                                 <option>on-failure</option> it will be
                                 restarted only when it exited with an
-                                exit code not equalling 0, or when
-                                terminated by a signal. If set to
+                                exit code not equalling 0, when
+                                terminated by a signal, when an
+                                operation times out or when the
+                                configured watchdog timeout is
+                                triggered. If set to
                                 <option>on-abort</option> it will be
                                 restarted only if it exits due to
                                 reception of an uncaught signal. If
                                 set to <option>always</option> the
                                 service will be restarted regardless
-                                whether it exited cleanly or not, or
+                                whether it exited cleanly or not,
                                 got terminated abnormally by a
-                                signal.</para></listitem>
+                                signal or hit a timeout.</para></listitem>
                         </varlistentry>
 
                         <varlistentry>
                                 accepted. If <option>all</option> all
                                 services updates from all members of
                                 the service's control group are
-                                accepted. This option must be set to
+                                accepted. This option should be set to
                                 open access to the notification socket
                                 when using
-                                <varname>Type=notify</varname> (see above).</para></listitem>
+                                <varname>Type=notify</varname> or
+                                <varname>WatchdogUsec=</varname> (see
+                                above). If those options are used but
+                                <varname>NotifyAccess=</varname> not
+                                configured it will be implicitly set
+                                to
+                                <option>main</option>.</para></listitem>
                         </varlistentry>
 
                         <varlistentry>
                                 <listitem><para>Specifies the name of
                                 the socket units this service shall
                                 inherit the sockets from when the
-                                service (ignoring the different suffix
-                                of course) is started. Normally it
+                                service is started. Normally it
                                 should not be necessary to use this
                                 setting as all sockets whose unit
                                 shares the same name as the service
+                                (ignoring the different suffix of course)
                                 are passed to the spawned
                                 process.</para>
 
                                 for details.</para></listitem>
                         </varlistentry>
 
+                        <varlistentry>
+                                <term><varname>StartLimitInterval=</varname></term>
+                                <term><varname>StartLimitBurst=</varname></term>
+
+                                <listitem><para>Configure service
+                                start rate limiting. By default
+                                services which are started more often
+                                than 5 times within 10s are not
+                                permitted to start any more times
+                                until the 10s interval ends. With
+                                these two options this rate limiting
+                                may be modified. Use
+                                <varname>StartLimitInterval=</varname>
+                                to configure the checking interval
+                                (defaults to 10s, set to 0 to disable
+                                any kind of rate limiting). Use
+                                <varname>StartLimitBurst=</varname> to
+                                configure how many starts per interval
+                                are allowed (defaults to 5). These
+                                configuration options are particularly
+                                useful in conjunction with
+                                <varname>Restart=</varname>.</para></listitem>
+                        </varlistentry>
+
+                        <varlistentry>
+                                <term><varname>StartLimitAction=</varname></term>
+
+                                <listitem><para>Configure the action
+                                to take if the rate limit configured
+                                with
+                                <varname>StartLimitInterval=</varname>
+                                and
+                                <varname>StartLimitBurst=</varname> is
+                                hit. Takes one of
+                                <option>none</option>,
+                                <option>reboot</option>,
+                                <option>reboot-force</option> or
+                                <option>reboot-immediate</option>. If
+                                <option>none</option> is set,
+                                hitting the rate limit will trigger no
+                                action besides that the start will not
+                                be
+                                permitted. <option>reboot</option>
+                                causes a reboot following the normal
+                                shutdown procedure (i.e. equivalent to
+                                <command>systemctl reboot</command>),
+                                <option>reboot-force</option> causes
+                                an forced reboot which will terminate
+                                all processes forcibly but should
+                                cause no dirty file systems on reboot
+                                (i.e. equivalent to <command>systemctl
+                                reboot -f</command>) and
+                                <option>reboot-immediate</option>
+                                causes immediate execution of the
+                                <citerefentry><refentrytitle>reboot</refentrytitle><manvolnum>2</manvolnum></citerefentry>
+                                system call, which might result in
+                                data loss.  Defaults to
+                                <option>none</option>.</para></listitem>
+                        </varlistentry>
+
                 </variablelist>
         </refsect1>