path: root/system/doc/design_principles/sup_princ.xml

<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">

<chapter>
  <header>
    <copyright>
      <year>1997</year><year>2016</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      Licensed under the Apache License, Version 2.0 (the "License");
      you may not use this file except in compliance with the License.
      You may obtain a copy of the License at
 
          http://www.apache.org/licenses/LICENSE-2.0

      Unless required by applicable law or agreed to in writing, software
      distributed under the License is distributed on an "AS IS" BASIS,
      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
      See the License for the specific language governing permissions and
      limitations under the License.
    
    </legalnotice>

    <title>Supervisor Behaviour</title>
    <prepared></prepared>
    <docno></docno>
    <date></date>
    <rev></rev>
    <file>sup_princ.xml</file>
  </header>
  <p>This section should be read with the
    <seealso marker="stdlib:supervisor">supervisor(3)</seealso> manual page
    in STDLIB, where all details about the supervisor
    behaviour is given.</p>

  <section>
    <title>Supervision Principles</title>
    <p>A supervisor is responsible for starting, stopping, and
      monitoring its child processes. The basic idea of a supervisor is
      that it is to keep its child processes alive by restarting them
      when necessary.</p>
    <p>Which child processes to start and monitor is specified by a
      list of <seealso marker="#spec">child specifications</seealso>.
      The child processes are started in the order specified by this
      list, and terminated in the reversed order.</p>
  </section>

  <section>
    <title>Example</title>
    <p>The callback module for a supervisor starting the server from
    <seealso marker="gen_server_concepts#ex">gen_server Behaviour</seealso>
    can look as follows:</p>
    <marker id="ex"></marker>
    <code type="none">
-module(ch_sup).
-behaviour(supervisor).

-export([start_link/0]).
-export([init/1]).

start_link() ->
    supervisor:start_link(ch_sup, []).

init(_Args) ->
    SupFlags = #{strategy => one_for_one, intensity => 1, period => 5},
    ChildSpecs = [#{id => ch3,
                    start => {ch3, start_link, []},
                    restart => permanent,
                    shutdown => brutal_kill,
                    type => worker,
                    modules => [cg3]}],
    {ok, {SupFlags, ChildSpecs}}.</code>
    <p>The <c>SupFlags</c> variable in the return value
      from <c>init/1</c> represents
      the <seealso marker="#flags">supervisor flags</seealso>.</p>
    <p>The <c>ChildSpecs</c> variable in the return value
      from <c>init/1</c> is a list of <seealso marker="#spec">child
      specifications</seealso>.</p>
    </section>

  <section>
    <marker id="flags"/>
    <title>Supervisor Flags</title>
    <p>This is the type definition for the supervisor flags:</p>
    <code type="none"><![CDATA[
sup_flags() = #{strategy => strategy(),         % optional
                intensity => non_neg_integer(), % optional
                period => pos_integer()}        % optional
    strategy() = one_for_all
               | one_for_one
               | rest_for_one
               | simple_one_for_one]]></code>
    <list type="bulleted">
      <item>
	<p><c>strategy</c> specifies
	  the <seealso marker="#strategy">restart
	  strategy</seealso>.</p>
      </item>
      <item>
	<p><c>intensity</c> and <c>period</c> specify
	  the <seealso marker="#max_intensity">maximum restart
	  intensity</seealso>.</p>
      </item>
    </list>
  </section>

  <section>
    <marker id="strategy"/>
    <title>Restart Strategy</title>
    <p> The restart strategy is specified by
      the <c>strategy</c> key in the supervisor flags map returned by
      the callback function <c>init</c>:</p>
    <code type="none">
SupFlags = #{strategy => Strategy, ...}</code>
    <p>The <c>strategy</c> key is optional in this map. If it is not
      given, it defaults to <c>one_for_one</c>.</p>

    <section>
      <title>one_for_one</title>
      <p>If a child process terminates, only that process is restarted.</p>
      <marker id="sup4"></marker>
      <image file="../design_principles/sup4.gif">
        <icaption>One_For_One Supervision</icaption>
      </image>
    </section>

    <section>
      <title>one_for_all</title>
      <p>If a child process terminates, all other child processes are
        terminated, and then all child processes, including
        the terminated one, are restarted.</p>
      <marker id="sup5"></marker>
      <image file="../design_principles/sup5.gif">
        <icaption>One_For_All Supervision</icaption>
      </image>
    </section>

    <section>
      <title>rest_for_one</title>
      <p>If a child process terminates, the rest of the child
        processes (that is, the child processes after the terminated
        process in start order) are terminated. Then the terminated
        child process and the rest of the child processes are restarted.</p>
    </section>

     <section>
      <title>simple_one_for_one</title>
      <p>See <seealso marker="#simple">simple-one-for-one
	  supervisors</seealso>.</p>
    </section>
  </section>

  <section>
    <marker id="max_intensity"></marker>
    <title>Maximum Restart Intensity</title>
    <p>The supervisors have a built-in mechanism to limit the number of
      restarts which can occur in a given time interval. This is
      specified by the two keys <c>intensity</c> and
      <c>period</c> in the supervisor flags map returned by the
      callback function <c>init</c>:</p>
    <code type="none">
SupFlags = #{intensity => MaxR, period => MaxT, ...}</code>
    <p>If more than <c>MaxR</c> number of restarts occur in the last
      <c>MaxT</c> seconds, the supervisor terminates all the child
      processes and then itself.
      The termination reason for the supervisor itself in that case will be
      <c>shutdown</c>.</p>
    <p>When the supervisor terminates, then the next higher-level
      supervisor takes some action. It either restarts the terminated
      supervisor or terminates itself.</p>
    <p>The intention of the restart mechanism is to prevent a situation
      where a process repeatedly dies for the same reason, only to be
      restarted again.</p>
    <p>The keys <c>intensity</c> and <c>period</c> are optional in the
      supervisor flags map. If they are not given, they default
      to <c>1</c> and <c>5</c>, respectively.</p>
    <section>
      <title>Tuning the intensity and period</title>
      <p>The default values are 1 restart per 5 seconds. This was chosen to
        be safe for most systems, even with deep supervision hierarchies,
        but you will probably want to tune the settings for your particular
        use case.</p>
      <p>First, the intensity decides how big bursts of restarts you want
        to tolerate. For example, you might want to accept a burst of at
        most 5 or 10 attempts, even within the same second, if it results
        in a successful restart.</p>
      <p>Second, you need to consider the sustained failure rate, if
        crashes keep happening but not often enough to make the supervisor
        give up. If you set intensity to 10 and set the period as low as 1,
        the supervisor will allow child processes to keep restarting up to
        10 times per second, forever, filling your logs with crash reports
        until someone intervenes manually.</p>
      <p>You should therefore set the period to be long enough that you can
        accept that the supervisor keeps going at that rate. For example,
        if you have picked an intensity value of 5, then setting the period
        to 30 seconds will give you at most one restart per 6 seconds for
        any longer period of time, which means that your logs won't fill up
        too quickly, and you will have a chance to observe the failures and
        apply a fix.</p>
      <p>These choices depend a lot on your problem domain. If you don't
        have real time monitoring and ability to fix problems quickly, for
        example in an embedded system, you might want to accept at most
        one restart per minute before the supervisor should give up and
        escalate to the next level to try to clear the error automatically.
        On the other hand, if it is more important that you keep trying
        even at a high failure rate, you might want a sustained rate of as
        much as 1-2 restarts per second.</p>
      <p>Avoiding common mistakes:</p>
      <list type="bulleted">
        <item>
          <p>Do not forget to consider the burst rate. If you set intensity
            to 1 and period to 6, it gives the same sustained error rate as
            5/30 or 10/60, but will not allow even 2 restart attempts in
            quick succession. This is probably not what you wanted.</p>
        </item>
        <item>
          <p>Do not set the period to a very high value if you want to
            tolerate bursts. If you set intensity to 5 and period to 3600
            (one hour), the supervisor will allow a short burst of 5
            restarts, but then gives up if it sees another single restart
            almost an hour later. You probably want to regard those crashes
            as separate incidents, so setting the period to 5 or 10 minutes
            will be more reasonable.</p>
        </item>
        <item>
          <p>If your application has multiple levels of supervision, then
            do not simply set the restart intensities to the same values on
            all levels. Keep in mind that the total number of restarts
            (before the top level supervisor gives up and terminates the
            application) will be the product of the intensity values of all
            the supervisors above the failing child process.</p>
          <p>For example, if the top level allows 10 restarts, and the next
            level also allows 10, a crashing child below that level will be
            restarted 100 times, which is probably excessive. Allowing at
            most 3 restarts for the top level supervisor might be a better
            choice in this case.</p>
        </item>
      </list>
    </section>
  </section>

  <section>
    <marker id="spec"></marker>
    <title>Child Specification</title>
    <p>The type definition for a child specification is as follows:</p>
    <code type="none"><![CDATA[
child_spec() = #{id => child_id(),       % mandatory
                 start => mfargs(),      % mandatory
                 restart => restart(),   % optional
                 shutdown => shutdown(), % optional
                 type => worker(),       % optional
                 modules => modules()}   % optional
    child_id() = term()
    mfargs() = {M :: module(), F :: atom(), A :: [term()]}
    modules() = [module()] | dynamic
    restart() = permanent | transient | temporary
    shutdown() = brutal_kill | timeout()
    worker() = worker | supervisor]]></code>
    <list type="bulleted">
      <item>
        <p><c>id</c> is used to identify the child
          specification internally by the supervisor.</p>
	<p>The <c>id</c> key is mandatory.</p>
	<p>Note that this identifier occasionally has been called
	  "name". As far as possible, the terms "identifier" or "id"
	  are now used but in order to keep backwards compatibility,
	  some occurences of "name" can still be found, for example
	  in error messages.</p>
      </item>
      <item>
        <p><c>start</c> defines the function call used to start
          the child process. It is a module-function-arguments tuple
          used as <c>apply(M, F, A)</c>.</p>
        <p>It is to be (or result in) a call to any of the following:</p>
	<list type="bulleted">
	  <item><c>supervisor:start_link</c></item>
	  <item><c>gen_server:start_link</c></item>
	  <item><c>gen_fsm:start_link</c></item>
	  <item><c>gen_statem:start_link</c></item>
	  <item><c>gen_event:start_link</c></item>
	  <item>A function compliant with these functions. For details,
	  see the <c>supervisor(3)</c> manual page.</item>
	</list>
	<p>The <c>start</c> key is mandatory.</p>
      </item>
      <item>
        <p><c>restart</c> defines when a terminated child process is to
          be restarted.</p>
        <list type="bulleted">
          <item>A <c>permanent</c> child process is always restarted.</item>
          <item>A <c>temporary</c> child process is never restarted
          (not even when the supervisor restart strategy
          is <c>rest_for_one</c> or <c>one_for_all</c> and a sibling
          death causes the temporary process to be terminated).</item>
          <item>A <c>transient</c> child process is restarted only if it
	  terminates abnormally, that is, with an exit reason other than
          <c>normal</c>, <c>shutdown</c>, or <c>{shutdown,Term}</c>.</item>
        </list>
	<p>The <c>restart</c> key is optional. If it is not given, the
	  default value <c>permanent</c> will be used.</p>
      </item>
      <item>
        <marker id="shutdown"></marker>
        <p><c>shutdown</c> defines how a child process is to be
          terminated.</p>
        <list type="bulleted">
          <item><c>brutal_kill</c> means that the child process is
           unconditionally terminated using <c>exit(Child, kill)</c>.</item>
          <item>An integer time-out value means that the supervisor tells
           the child process to terminate by calling
          <c>exit(Child, shutdown)</c> and then waits for an exit
           signal back. If no exit signal is received within
           the specified time, the child process is unconditionally
           terminated using <c>exit(Child, kill)</c>.</item>
          <item>If the child process is another supervisor, it is to be
           set to <c>infinity</c> to give the subtree enough time to
           shut down. It is also allowed to set it to <c>infinity</c>,
          if the child process is a worker. See the warning below:</item>
        </list>
        <warning>
          <p>Be careful when setting the shutdown time to
          <c>infinity</c> when the child process is a worker. Because, in this
          situation, the termination of the supervision tree depends on the
          child process; it must be implemented in a safe way and its cleanup
          procedure must always return.</p>
        </warning>
	<p>The <c>shutdown</c> key is optional. If it is not given,
	  and the child is of type <c>worker</c>, the default value
	  <c>5000</c> will be used; if the child is of type
	  <c>supervisor</c>, the default value <c>infinity</c> will be
	  used.</p>
      </item>
      <item>
        <p><c>type</c> specifies if the child process is a supervisor or
          a worker.</p>
	<p>The <c>type</c> key is optional. If it is not given, the
	  default value <c>worker</c> will be used.</p>
      </item>
      <item>
        <p><c>modules</c> are to be a list with one element
          <c>[Module]</c>, where <c>Module</c> is the name of
          the callback module, if the child process is a supervisor,
          gen_server, gen_fsm or gen_statem.
	  If the child process is a gen_event,
          the value shall be <c>dynamic</c>.</p>
        <p>This information is used by the release handler during
          upgrades and downgrades, see
          <seealso marker="release_handling">Release Handling</seealso>.</p>
	<p>The <c>modules</c> key is optional. If it is not given, it
	  defaults to <c>[M]</c>, where <c>M</c> comes from the
	  child's start <c>{M,F,A}</c>.</p>
      </item>
    </list>
    <p><em>Example:</em> The child specification to start the server
    <c>ch3</c> in the previous example look as follows:</p>
    <code type="none">
#{id => ch3,
  start => {ch3, start_link, []},
  restart => permanent,
  shutdown => brutal_kill,
  type => worker,
  modules => [ch3]}</code>
    <p>or simplified, relying on the default values:</p>
    <code type="none">
#{id => ch3,
  start => {ch3, start_link, []}
  shutdown => brutal_kill}</code>
    <p>Example: A child specification to start the event manager from
      the chapter about
      <seealso marker="events#mgr">gen_event</seealso>:</p>
    <code type="none">
#{id => error_man,
  start => {gen_event, start_link, [{local, error_man}]},
  modules => dynamic}</code>
    <p>Both server and event manager are registered processes which
      can be expected to be always accessible. Thus they are
      specified to be <c>permanent</c>.</p>
    <p><c>ch3</c> does not need to do any cleaning up before
      termination. Thus, no shutdown time is needed, but
      <c>brutal_kill</c> is sufficient. <c>error_man</c> can
      need some time for the event handlers to clean up, thus
      the shutdown time is set to 5000 ms (which is the default
      value).</p>
    <p>Example: A child specification to start another supervisor:</p>
    <code type="none">
#{id => sup,
  start => {sup, start_link, []},
  restart => transient,
  type => supervisor} % will cause default shutdown=>infinity</code>
  </section>

  <section>
    <marker id="super_tree"></marker>
    <title>Starting a Supervisor</title>
    <p>In the previous example, the supervisor is started by calling
      <c>ch_sup:start_link()</c>:</p>
    <code type="none">
start_link() ->
    supervisor:start_link(ch_sup, []).</code>
    <p><c>ch_sup:start_link</c> calls function
    <c>supervisor:start_link/2</c>, which spawns and links to a new
    process, a supervisor.</p>
    <list type="bulleted">
      <item>The first argument, <c>ch_sup</c>, is the name of
       the callback module, that is, the module where the <c>init</c>
       callback function is located.</item>
      <item>The second argument, <c>[]</c>, is a term that is passed
       as is to
       the callback function <c>init</c>. Here, <c>init</c> does not
       need any indata and ignores the argument.</item>
    </list>
    <p>In this case, the supervisor is not registered. Instead its pid
      must be used. A name can be specified by calling
      <c>supervisor:start_link({local, Name}, Module, Args)</c> or
      <c>supervisor:start_link({global, Name}, Module, Args)</c>.</p>
    <p>The new supervisor process calls the callback function
      <c>ch_sup:init([])</c>. <c>init</c> shall return
      <c>{ok, {SupFlags, ChildSpecs}}</c>:</p>
    <code type="none">
init(_Args) ->
    SupFlags = #{},
    ChildSpecs = [#{id => ch3,
                    start => {ch3, start_link, []},
                    shutdown => brutal_kill}],
    {ok, {SupFlags, ChildSpecs}}.</code>
    <p>The supervisor then starts all its child processes according to
      the child specifications in the start specification. In this case
      there is one child process, <c>ch3</c>.</p>
    <p><c>supervisor:start_link</c> is synchronous. It does
      not return until all child processes have been started.</p>
  </section>

  <section>
    <title>Adding a Child Process</title>
    <p>In addition to the static supervision tree, dynamic child
    processes can be added to an existing supervisor with the following
    call:</p>
    <code type="none">
supervisor:start_child(Sup, ChildSpec)</code>
    <p><c>Sup</c> is the pid, or name, of the supervisor.
    <c>ChildSpec</c> is a
    <seealso marker="#spec">child specification</seealso>.</p>
    <p>Child processes added using <c>start_child/2</c> behave in
      the same way as the other child processes, with the an important
      exception: if a supervisor dies and is recreated, then
      all child processes that were dynamically added to the supervisor
      are lost.</p>
  </section>

  <section>
    <title>Stopping a Child Process</title>
    <p>Any child process, static or dynamic, can be stopped in
      accordance with the shutdown specification:</p>
    <code type="none">
supervisor:terminate_child(Sup, Id)</code>
    <p>The child specification for a stopped child process is deleted
      with the following call:</p>
    <code type="none">
supervisor:delete_child(Sup, Id)</code>
    <p><c>Sup</c> is the pid, or name, of the supervisor.
      <c>Id</c> is the value associated with the <c>id</c> key in
      the <seealso marker="#spec">child specification</seealso>.</p>
    <p>As with dynamically added child processes, the effects of
      deleting a static child process is lost if the supervisor itself
      restarts.</p>
  </section>

  <section>
    <marker id="simple"/>
    <title>Simplified one_for_one Supervisors</title>
    <p>A supervisor with restart strategy <c>simple_one_for_one</c> is
      a simplified <c>one_for_one</c> supervisor, where all child
      processes are dynamically added instances of the same process.</p>
    <p>The following is an example of a callback module for a
    <c>simple_one_for_one</c> supervisor:</p>
    <code type="none">
-module(simple_sup).
-behaviour(supervisor).

-export([start_link/0]).
-export([init/1]).

start_link() ->
    supervisor:start_link(simple_sup, []).

init(_Args) ->
    SupFlags = #{strategy => simple_one_for_one,
                 intensity => 0,
                 period => 1},
    ChildSpecs = [#{id => call,
                    start => {call, start_link, []},
                    shutdown => brutal_kill}],
    {ok, {SupFlags, ChildSpecs}}.</code>
    <p>When started, the supervisor does not start any child processes.
      Instead, all child processes are added dynamically by calling:</p>
    <code type="none">
supervisor:start_child(Sup, List)</code>
    <p><c>Sup</c> is the pid, or name, of the supervisor.
      <c>List</c> is an arbitrary list of terms, which are added to
      the list of arguments specified in the child specification. If
      the start function is specified as <c>{M, F, A}</c>,
      the child process is started by calling
      <c>apply(M, F, A++List)</c>.</p>
    <p>For example, adding a child to <c>simple_sup</c> above:</p>
    <code type="none">
supervisor:start_child(Pid, [id1])</code>
    <p>The result is that the child process is started by calling
      <c>apply(call, start_link, []++[id1])</c>, or actually:</p>
    <code type="none">
call:start_link(id1)</code>
    <p>A child under a <c>simple_one_for_one</c> supervisor can be
    terminated with the following:</p>
    <code type="none">
supervisor:terminate_child(Sup, Pid)</code>
    <p><c>Sup</c> is the pid, or name, of the supervisor and
    <c>Pid</c> is the pid of the child.</p>
    <p>Because a <c>simple_one_for_one</c> supervisor can have many
      children, it shuts them all down asynchronously. This means that
      the children will do their cleanup in parallel and therefore the
      order in which they are stopped is not defined.</p>
  </section>

  <section>
    <title>Stopping</title>
    <p>Since the supervisor is part of a supervision tree, it is
      automatically terminated by its supervisor. When asked to
      shut down, it terminates all child processes in reversed start
      order according to the respective shutdown specifications, and
      then terminates itself.</p>
  </section>
</chapter>