A behaviour module for implementing supervisors.

A supervisor is a process which supervises other processes, which we refer to as child processes. Supervisors are used to build a hierarchical process structure called a supervision tree. Supervision trees provide fault-tolerance and encapsulate how our applications start and shutdown.

A supervisor may be started directly with a list of child specifications via start_link/2 or you may define a module-based supervisor that implements the required callbacks. The sections below use start_link/2 to start supervisors in most examples, but it also includes a specific section on module-based ones.

Examples

In order to start a supervisor, we need to first define a child process that will be supervised. As an example, we will define a GenServer, a generic server, that keeps a counter. Other processes can then send messages to this process to read the counter and bump its value.

Disclaimer

In practice you would not define a counter as a GenServer. Instead, if you need a counter, you would pass it around as inputs and outputs to the functions that need it. The reason we picked a counter in this example is due to its simplicity, as it allows us to focus on how supervisors work.

defmodule Counter do
  use GenServer

  def start_link(arg) when is_integer(arg) do
    GenServer.start_link(__MODULE__, arg, name: __MODULE__)
  end

  ## Callbacks

  @impl true
  def init(counter) do
    {:ok, counter}
  end

  @impl true
  def handle_call(:get, _from, counter) do
    {:reply, counter, counter}
  end

  def handle_call({:bump, value}, _from, counter) do
    {:reply, counter, counter + value}
  end
end

The Counter receives an argument on start_link. This argument is passed to the init/1 callback which becomes the initial value of the counter. Our counter handles two operations (known as calls): :get, to get the current counter value, and :bump, that bumps the counter by the given value and returns the old counter.

We can now start a supervisor that will start and supervise our counter process. The first step is to define a list of child specifications that control how each child behaves. Each child specification is a map, as shown below:

children = [
  # The Counter is a child started via Counter.start_link(0)
  %{
    id: Counter,
    start: {Counter, :start_link, [0]}
  }
]

# Now we start the supervisor with the children and a strategy
{:ok, pid} = Supervisor.start_link(children, strategy: :one_for_one)

# After started, we can query the supervisor for information
Supervisor.count_children(pid)
#=> %{active: 1, specs: 1, supervisors: 0, workers: 1}

Note that when starting the GenServer, we are registering it with name Counter via the name: __MODULE__ option. This allows us to call it directly and get its value:

GenServer.call(Counter, :get)
#=> 0

GenServer.call(Counter, {:bump, 3})
#=> 0

GenServer.call(Counter, :get)
#=> 3

However, there is a bug in our counter server. If we call :bump with a non-numeric value, it is going to crash:

GenServer.call(Counter, {:bump, "oops"})
** (exit) exited in: GenServer.call(Counter, {:bump, "oops"}, 5000)

Luckily, since the server is being supervised by a supervisor, the supervisor will automatically start a new one, reset back to its initial value of 0:

GenServer.call(Counter, :get)
#=> 0

Supervisors support different strategies; in the example above, we have chosen :one_for_one. Furthermore, each supervisor can have many workers and/or supervisors as children, with each one having its own configuration (as outlined in the "Child specification" section).

The rest of this document will cover how child processes are specified, how they can be started and stopped, different supervision strategies and more.

Child specification

The child specification describes how the supervisor starts, shuts down, and restarts child processes.

The child specification is a map containing up to 6 elements. The first two keys in the following list are required, and the remaining ones are optional:

• :id - any term used to identify the child specification internally by the supervisor; defaults to the given module. This key is required. For supervisors, in the case of conflicting :id values, the supervisor will refuse to initialize and require explicit IDs. This is not the case for dynamic supervisors though.

• :start - a tuple with the module-function-args to be invoked to start the child process. This key is required.

• :restart - an atom that defines when a terminated child process should be restarted (see the "Restart values" section below). This key is optional and defaults to :permanent.

• :shutdown - an integer or atom that defines how a child process should be terminated (see the "Shutdown values" section below). This key is optional and defaults to 5_000 if the type is :worker or :infinity if the type is :supervisor.

• :type - specifies that the child process is a :worker or a :supervisor. This key is optional and defaults to :worker.

• :modules - a list of modules used by hot code upgrade mechanisms to determine which processes are using certain modules. It is typically set to the callback module of behaviours like GenServer, Supervisor, and such. It is set automatically based on the :start value and it is rarely changed in practice.

• :significant - a boolean indicating if the child process should be considered significant with regard to automatic shutdown. Only :transient and :temporary child processes can be marked as significant. This key is optional and defaults to false. See section "Automatic shutdown" below for more details.

Let's understand what the :shutdown and :restart options control.

Shutdown values (:shutdown)

The following shutdown values are supported in the :shutdown option:

• :brutal_kill - the child process is unconditionally and immediately terminated using Process.exit(child, :kill).

• any integer >= 0 - the amount of time in milliseconds that the supervisor will wait for its children to terminate after emitting a Process.exit(child, :shutdown) signal. If the child process is not trapping exits, the initial :shutdown signal will terminate the child process immediately. If the child process is trapping exits, it has the given amount of time to terminate. If it doesn't terminate within the specified time, the child process is unconditionally terminated by the supervisor via Process.exit(child, :kill).

• :infinity - works as an integer except the supervisor will wait indefinitely for the child to terminate. If the child process is a supervisor, the recommended value is :infinity to give the supervisor and its children enough time to shut down. This option can be used with regular workers but doing so is discouraged and requires extreme care. If not used carefully, the child process will never terminate, preventing your application from terminating as well.

Restart values (:restart)

The :restart option controls what the supervisor should consider to be a successful termination or not. If the termination is successful, the supervisor won't restart the child. If the child process crashed, the supervisor will start a new one.

The following restart values are supported in the :restart option:

• :permanent - the child process is always restarted.

• :temporary - the child process is never restarted, regardless of the supervision strategy: any termination (even abnormal) is considered successful.

• :transient - the child process is restarted only if it terminates abnormally, i.e., with an exit reason other than :normal, :shutdown, or {:shutdown, term}.

For a more complete understanding of the exit reasons and their impact, see the "Exit reasons and restarts" section.

child_spec/1 function

When starting a supervisor, we may pass a list of child specifications. Those specifications are maps that tell how the supervisor should start, stop and restart each of its children:

%{
  id: Counter,
  start: {Counter, :start_link, [0]}
}

The map above defines a child with :id of Counter that is started by calling Counter.start_link(0).

However, defining the child specification for each child as a map can be quite error prone, as we may change the Counter implementation and forget to update its specification. That's why Elixir allows you to pass a tuple with the module name and the start_link argument instead of the specification:

children = [
  {Counter, 0}
]

The supervisor will then invoke Counter.child_spec(0) to retrieve a child specification. Now the Counter module is responsible for building its own specification, for example, we could write:

def child_spec(arg) do
  %{
    id: Counter,
    start: {Counter, :start_link, [arg]}
  }
end

Then the supervisor will call Counter.start_link(arg) to start the child process. This flow is summarized in the diagram below. Caller is a process which spawns the Supervisor process. The Supervisor then proceeds to call your code (Module) to spawn its child process:

sequenceDiagram
    participant C as Caller (Process)
    participant S as Supervisor (Process)
    participant M as Module (Code)

    note right of C: child is a {module, arg} specification
    C->>+S: Supervisor.start_link([child])
    S-->>+M: module.child_spec(arg)
    M-->>-S: %{id: term, start: {module, :start_link, [arg]}}
    S-->>+M: module.start_link(arg)
    M->>M: Spawns child process (child_pid)
    M-->>-S: {:ok, child_pid} | :ignore | {:error, reason}
    S->>-C: {:ok, supervisor_pid} | {:error, reason}

Luckily for us, use GenServer already defines a Counter.child_spec/1 exactly like above, so you don't need to write the definition above yourself. If you want to customize the automatically generated child_spec/1 function, you can pass the options directly to use GenServer:

use GenServer, restart: :transient

Finally, note it is also possible to simply pass the Counter module as a child:

children = [
  Counter
]

When only the module name is given, it is equivalent to {Counter, []}, which in our case would be invalid, which is why we always pass the initial counter explicitly.

By replacing the child specification with {Counter, 0}, we keep it encapsulated in the Counter module. We could now share our Counter implementation with other developers and they can add it directly to their supervision tree without worrying about the low-level details of the counter.

Overall, a child specification can be one of the following:

• a map representing the child specification itself - as outlined in the "Child specification" section

• a tuple with a module as first element and the start argument as second - such as {Counter, 0}. In this case, Counter.child_spec(0) is called to retrieve the child specification

• a module - such as Counter. In this case, Counter.child_spec([]) would be called, which is invalid for the counter, but it is useful in many other cases, especially when you want to pass a list of options to the child process

If you need to convert a {module, arg} tuple or a module child specification to a child specification or modify a child specification itself, you can use the Supervisor.child_spec/2 function. For example, to run the counter with a different :id and a :shutdown value of 10 seconds (10_000 milliseconds):

children = [
  Supervisor.child_spec({Counter, 0}, id: MyCounter, shutdown: 10_000)
]

Supervisor strategies and options

So far we have started the supervisor passing a single child as a tuple as well as a strategy called :one_for_one:

children = [
  {Counter, 0}
]

Supervisor.start_link(children, strategy: :one_for_one)

The first argument given to start_link/2 is a list of child specifications as defined in the "child_spec/1" section above.

The second argument is a keyword list of options:

• :strategy - the supervision strategy option. It can be either :one_for_one, :rest_for_one or :one_for_all. Required. See the "Strategies" section.

• :max_restarts - the maximum number of restarts allowed in a time frame. Defaults to 3.

• :max_seconds - the time frame in which :max_restarts applies. Defaults to 5.

• :auto_shutdown - the automatic shutdown option. It can be :never, :any_significant, or :all_significant. Optional. See the "Automatic shutdown" section.

• :name - a name to register the supervisor process. Supported values are explained in the "Name registration" section in the documentation for GenServer. Optional.

Strategies

Supervisors support different supervision strategies (through the :strategy option, as seen above):

• :one_for_one - if a child process terminates, only that process is restarted.

• :one_for_all - if a child process terminates, all other child processes are terminated and then all child processes (including the terminated one) are restarted.

• :rest_for_one - if a child process terminates, the terminated child process and the rest of the children started after it, are terminated and restarted.

In the above, process termination refers to unsuccessful termination, which is determined by the :restart option.

To efficiently supervise children started dynamically, see DynamicSupervisor.

Automatic shutdown

Supervisors have the ability to automatically shut themselves down when child processes marked as :significant exit.

Supervisors support different automatic shutdown options (through the :auto_shutdown option, as seen above):

• :never - this is the default, automatic shutdown is disabled.

• :any_significant - if any significant child process exits, the supervisor will automatically shut down its children, then itself.

• :all_significant - when all significant child processes have exited, the supervisor will automatically shut down its children, then itself.

Only :transient and :temporary child processes can be marked as significant, and this configuration affects the behavior. Significant :transient child processes must exit normally for automatic shutdown to be considered, where :temporary child processes may exit for any reason.

Name registration

A supervisor is bound to the same name registration rules as a GenServer. Read more about these rules in the documentation for GenServer.

Module-based supervisors

In the example so far, the supervisor was started by passing the supervision structure to start_link/2. However, supervisors can also be created by explicitly defining a supervision module:

defmodule MyApp.Supervisor do
  # Automatically defines child_spec/1
  use Supervisor

  def start_link(init_arg) do
    Supervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
  end

  @impl true
  def init(_init_arg) do
    children = [
      {Counter, 0}
    ]

    Supervisor.init(children, strategy: :one_for_one)
  end
end

The difference between the two approaches is that a module-based supervisor gives you more direct control over how the supervisor is initialized. Instead of calling Supervisor.start_link/2 with a list of child specifications that are implicitly initialized for us, we must explicitly initialize the children by calling Supervisor.init/2 inside its init/1 callback. Supervisor.init/2 accepts the same :strategy, :max_restarts, and :max_seconds options as start_link/2.

use Supervisor

When you use Supervisor, the Supervisor module will set @behaviour Supervisor and define a child_spec/1 function, so your module can be used as a child in a supervision tree.

use Supervisor also defines a child_spec/1 function which allows us to run MyApp.Supervisor as a child of another supervisor or at the top of your supervision tree as:

children = [
  MyApp.Supervisor
]

Supervisor.start_link(children, strategy: :one_for_one)

A general guideline is to use the supervisor without a callback module only at the top of your supervision tree, generally in the Application.start/2 callback. We recommend using module-based supervisors for any other supervisor in your application, so they can run as a child of another supervisor in the tree. The child_spec/1 generated automatically by Supervisor can be customized with the following options:

• :id - the child specification identifier, defaults to the current module
• :restart - when the supervisor should be restarted, defaults to :permanent

The @doc annotation immediately preceding use Supervisor will be attached to the generated child_spec/1 function.

Start and shutdown

When the supervisor starts, it traverses all child specifications and then starts each child in the order they are defined. This is done by calling the function defined under the :start key in the child specification and typically defaults to start_link/1.

The start_link/1 (or a custom) is then called for each child process. The start_link/1 function must return {:ok, pid} where pid is the process identifier of a new process that is linked to the supervisor. The child process usually starts its work by executing the init/1 callback. Generally speaking, the init callback is where we initialize and configure the child process.

The shutdown process happens in reverse order.

When a supervisor shuts down, it terminates all children in the opposite order they are listed. The termination happens by sending a shutdown exit signal, via Process.exit(child_pid, :shutdown), to the child process and then awaiting for a time interval for the child process to terminate. This interval defaults to 5000 milliseconds. If the child process does not terminate in this interval, the supervisor abruptly terminates the child with reason :kill. The shutdown time can be configured in the child specification which is fully detailed in the next section.

If the child process is not trapping exits, it will shutdown immediately when it receives the first exit signal. If the child process is trapping exits, then the terminate callback is invoked, and the child process must terminate in a reasonable time interval before being abruptly terminated by the supervisor.

In other words, if it is important that a process cleans after itself when your application or the supervision tree is shutting down, then this process must trap exits and its child specification should specify the proper :shutdown value, ensuring it terminates within a reasonable interval.

Exit reasons and restarts

A supervisor restarts a child process depending on its :restart configuration. For example, when :restart is set to :transient, the supervisor does not restart the child in case it exits with reason :normal, :shutdown or {:shutdown, term}.

Those exits also impact logging. By default, behaviours such as GenServers do not emit error logs when the exit reason is :normal, :shutdown or {:shutdown, term}.

So one may ask: which exit reason should I choose? There are three options:

• :normal - in such cases, the exit won't be logged, there is no restart in transient mode, and linked processes do not exit

• :shutdown or {:shutdown, term} - in such cases, the exit won't be logged, there is no restart in transient mode, and linked processes exit with the same reason unless they're trapping exits

• any other term - in such cases, the exit will be logged, there are restarts in transient mode, and linked processes exit with the same reason unless they're trapping exits

Generally speaking, if you are exiting for expected reasons, you want to use :shutdown or {:shutdown, term}.

Note that the supervisor that reaches maximum restart intensity will exit with :shutdown reason. In this case the supervisor will only be restarted if its child specification was defined with the :restart option set to :permanent (the default).