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

<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">

<chapter>
  <header>
    <copyright>
      <year>1997</year><year>2011</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      The contents of this file are subject to the Erlang Public License,
      Version 1.1, (the "License"); you may not use this file except in
      compliance with the License. You should have received a copy of the
      Erlang Public License along with this software. If not, it can be
      retrieved online at http://www.erlang.org/.
    
      Software distributed under the License is distributed on an "AS IS"
      basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
      the License for the specific language governing rights 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 in conjunction with
    <c>supervisor(3)</c>, 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 should 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
      the <seealso marker="gen_server_concepts#ex">gen_server chapter</seealso>
      could look like this:</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) ->
    {ok, {{one_for_one, 1, 60},
          [{ch3, {ch3, start_link, []},
            permanent, brutal_kill, worker, [ch3]}]}}.</code>
    <p><c>one_for_one</c> is the <seealso marker="#strategy">restart strategy</seealso>.</p>
    <p>1 and 60 defines the <seealso marker="#frequency">maximum restart frequency</seealso>.</p>
    <p>The tuple <c>{ch3, ...}</c> is a <seealso marker="#spec">child specification</seealso>.</p>
  </section>

  <section>
    <marker id="strategy"></marker>
    <title>Restart Strategy</title>

    <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 -- i.e. 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>

  <section>
    <marker id="frequency"></marker>
    <title>Maximum Restart Frequency</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
      determined by the values of the two parameters <c>MaxR</c> and
      <c>MaxT</c> in the start specification returned by the callback
      function <c>init</c>:</p>
    <code type="none">
init(...) ->
    {ok, {{RestartStrategy, MaxR, MaxT},
          [ChildSpec, ...]}}.</code>
    <p>If more than <c>MaxR</c> number of restarts occur in the last
      <c>MaxT</c> seconds, then the supervisor terminates all the child
      processes and then itself.</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>
  </section>

  <section>
    <marker id="spec"></marker>
    <title>Child Specification</title>
    <p>This is the type definition for a child specification:</p>
    <code type="none"><![CDATA[
{Id, StartFunc, Restart, Shutdown, Type, Modules}
    Id = term()
    StartFunc = {M, F, A}
        M = F = atom()
        A = [term()]
    Restart = permanent | transient | temporary
    Shutdown = brutal_kill | integer() &gt;=0 | infinity
    Type = worker | supervisor
    Modules = [Module] | dynamic
        Module = atom()]]></code>
    <list type="bulleted">
      <item>
        <p><c>Id</c> is a name that is used to identify the child
          specification internally by the supervisor.</p>
      </item>
      <item>
        <p><c>StartFunc</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 should be (or result in) a call to
          <c>supervisor:start_link</c>, <c>gen_server:start_link</c>,
          <c>gen_fsm:start_link</c> or <c>gen_event:start_link</c>.
          (Or a function compliant with these functions, see
          <c>supervisor(3)</c> for details.</p>
      </item>
      <item>
        <p><c>Restart</c> defines when a terminated child process should
          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.</item>
          <item>A <c>transient</c> child process is restarted only if it
           terminates abnormally, i.e. with another exit reason than
          <c>normal</c>.</item>
        </list>
      </item>
      <item>
        <marker id="shutdown"></marker>
        <p><c>Shutdown</c> defines how a child process should be
          terminated.</p>
        <list type="bulleted">
          <item><c>brutal_kill</c> means the child process is
           unconditionally terminated using <c>exit(Child, kill)</c>.</item>
          <item>An integer timeout 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 should be
           set to <c>infinity</c> to give the subtree enough time to
           shutdown. It is also allowed to set it to <c>infinity</c>, if the
           child process is a worker.</item>
        </list>
      </item>
      <item>
        <p><c>Type</c> specifies if the child process is a supervisor or
          a worker.</p>
      </item>
      <item>
        <p><c>Modules</c> should 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 or gen_fsm. If the child process is a gen_event,
          <c>Modules</c> should 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>
      </item>
    </list>
    <p>Example: The child specification to start the server <c>ch3</c>
      in the example above looks like:</p>
    <code type="none">
{ch3,
 {ch3, start_link, []},
 permanent, brutal_kill, worker, [ch3]}</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">
{error_man,
 {gen_event, start_link, [{local, error_man}]},
 permanent, 5000, worker, dynamic}</code>
    <p>Both the server and event manager are registered processes which
      can be expected to be accessible at all times, 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> should be sufficient. <c>error_man</c> may
      need some time for the event handlers to clean up, thus
      <c>Shutdown</c> is set to 5000 ms.</p>
    <p>Example: A child specification to start another supervisor:</p>
    <code type="none">
{sup,
 {sup, start_link, []},
 transient, infinity, supervisor, [sup]}</code>
  </section>

  <section>
    <marker id="super_tree"></marker>
    <title>Starting a Supervisor</title>
    <p>In the example above, 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 the function
      <c>supervisor:start_link/2</c>. This function 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, [], is a term which 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> is expected to return
      <c>{ok, StartSpec}</c>:</p>
    <code type="none">
init(_Args) ->
    {ok, {{one_for_one, 1, 60},
          [{ch3, {ch3, start_link, []},
            permanent, brutal_kill, worker, [ch3]}]}}.</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>Note that <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, we can also add
      dynamic child processes 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 manner as the other child processes, with the following
      important exception: If a supervisor dies and is re-created, then
      all child processes which were dynamically added to the supervisor
      will be 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 id specified 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>
    <title>Simple-One-For-One Supervisors</title>
    <p>A supervisor with restart strategy <c>simple_one_for_one</c> is
      a simplified one_for_one supervisor, where all child processes are
      dynamically added instances of the same process.</p>
    <p>Example of a callback module for a simple_one_for_one 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) ->
    {ok, {{simple_one_for_one, 0, 1},
          [{call, {call, start_link, []},
            temporary, brutal_kill, worker, [call]}]}}.</code>
    <p>When started, the supervisor will 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 will be added to
      the list of arguments specified in the child specification. If
      the start function is specified as <c>{M, F, A}</c>, then
      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>results in the child process being 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</p>
    <code type="none">
supervisor:terminate_child(Sup, Pid)</code>
    <p>where <c>Sup</c> is the pid, or name, of the supervisor and
    <c>Pid</c> is the pid of the child.</p>
  </section>

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