A behaviour module for implementing the server of a client-server relation.

A GenServer is a process like any other Elixir process and it can be used to keep state, execute code asynchronously and so on. The advantage of using a generic server process (GenServer) implemented using this module is that it will have a standard set of interface functions and include functionality for tracing and error reporting. It will also fit into a supervision tree.

graph BT
    C(Client #3) ~~~ B(Client #2) ~~~ A(Client #1)
    A & B & C -->|request| GenServer
    GenServer -.->|reply| A & B & C

Example

The GenServer behaviour abstracts the common client-server interaction. Developers are only required to implement the callbacks and functionality they are interested in.

Let's start with a code example and then explore the available callbacks. Imagine we want to implement a service with a GenServer that works like a stack, allowing us to push and pop elements. We'll customize a generic GenServer with our own module by implementing three callbacks.

init/1 transforms our initial argument to the initial state for the GenServer. handle_call/3 fires when the server receives a synchronous pop message, popping an element from the stack and returning it to the user. handle_cast/2 will fire when the server receives an asynchronous push message, pushing an element onto the stack:

defmodule Stack do
  use GenServer

  # Callbacks

  @impl true
  def init(elements) do
    initial_state = String.split(elements, ",", trim: true)
    {:ok, initial_state}
  end

  @impl true
  def handle_call(:pop, _from, state) do
    [to_caller | new_state] = state
    {:reply, to_caller, new_state}
  end

  @impl true
  def handle_cast({:push, element}, state) do
    new_state = [element | state]
    {:noreply, new_state}
  end
end

We leave the process machinery of startup, message passing, and the message loop to the GenServer behaviour and focus only on the stack implementation. We can now use the GenServer API to interact with the service by creating a process and sending it messages:

# Start the server
{:ok, pid} = GenServer.start_link(Stack, "hello,world")

# This is the client
GenServer.call(pid, :pop)
#=> "hello"

GenServer.cast(pid, {:push, "elixir"})
#=> :ok

GenServer.call(pid, :pop)
#=> "elixir"

We start our Stack by calling start_link/2, passing the module with the server implementation and its initial argument with a comma-separated list of elements. The GenServer behaviour calls the init/1 callback to establish the initial GenServer state. From this point on, the GenServer has control so we interact with it by sending two types of messages on the client. call messages expect a reply from the server (and are therefore synchronous) while cast messages do not.

Each call to GenServer.call/3 results in a message that must be handled by the handle_call/3 callback in the GenServer. A cast/2 message must be handled by handle_cast/2. GenServer supports 8 callbacks, but only init/1 is required.

Client / Server APIs

Although in the example above we have used GenServer.start_link/3 and friends to directly start and communicate with the server, most of the time we don't call the GenServer functions directly. Instead, we wrap the calls in new functions representing the public API of the server. These thin wrappers are called the client API.

Here is a better implementation of our Stack module:

defmodule Stack do
  use GenServer

  # Client

  def start_link(default) when is_binary(default) do
    GenServer.start_link(__MODULE__, default)
  end

  def push(pid, element) do
    GenServer.cast(pid, {:push, element})
  end

  def pop(pid) do
    GenServer.call(pid, :pop)
  end

  # Server (callbacks)

  @impl true
  def init(elements) do
    initial_state = String.split(elements, ",", trim: true)
    {:ok, initial_state}
  end

  @impl true
  def handle_call(:pop, _from, state) do
    [to_caller | new_state] = state
    {:reply, to_caller, new_state}
  end

  @impl true
  def handle_cast({:push, element}, state) do
    new_state = [element | state]
    {:noreply, new_state}
  end
end

In practice, it is common to have both server and client functions in the same module. If the server and/or client implementations are growing complex, you may want to have them in different modules.

The following diagram summarizes the interactions between client and server. Both Client and Server are processes and communication happens via messages (continuous line). The Server <-> Module interaction happens when the GenServer process calls your code (dotted lines):

sequenceDiagram
    participant C as Client (Process)
    participant S as Server (Process)
    participant M as Module (Code)

    note right of C: Typically started by a supervisor
    C->>+S: GenServer.start_link(module, arg, options)
    S-->>+M: init(arg)
    M-->>-S: {:ok, state} | :ignore | {:error, reason}
    S->>-C: {:ok, pid} | :ignore | {:error, reason}

    note right of C: call is synchronous
    C->>+S: GenServer.call(pid, message)
    S-->>+M: handle_call(message, from, state)
    M-->>-S: {:reply, reply, state} | {:stop, reason, reply, state}
    S->>-C: reply

    note right of C: cast is asynchronous
    C-)S: GenServer.cast(pid, message)
    S-->>+M: handle_cast(message, state)
    M-->>-S: {:noreply, state} | {:stop, reason, state}

    note right of C: send is asynchronous
    C-)S: Kernel.send(pid, message)
    S-->>+M: handle_info(message, state)
    M-->>-S: {:noreply, state} | {:stop, reason, state}

How to supervise

A GenServer is most commonly started under a supervision tree. When we invoke use GenServer, it automatically defines a child_spec/1 function that allows us to start the Stack directly under a supervisor. To start a default stack of ["hello", "world"] under a supervisor, we can do:

children = [
  {Stack, "hello,world"}
]

Supervisor.start_link(children, strategy: :one_for_all)

Note that specifying a module MyServer would be the same as specifying the tuple {MyServer, []}.

use GenServer also accepts a list of options which configures the child specification and therefore how it runs under a supervisor. The generated child_spec/1 can be customized with the following options:

• :id - the child specification identifier, defaults to the current module
• :restart - when the child should be restarted, defaults to :permanent
• :shutdown - how to shut down the child, either immediately or by giving it time to shut down

For example:

use GenServer, restart: :transient, shutdown: 10_000

See the "Child specification" section in the Supervisor module for more detailed information. The @doc annotation immediately preceding use GenServer will be attached to the generated child_spec/1 function.

When stopping the GenServer, for example by returning a {:stop, reason, new_state} tuple from a callback, the exit reason is used by the supervisor to determine whether the GenServer needs to be restarted. See the "Exit reasons and restarts" section in the Supervisor module.

Name registration

Both start_link/3 and start/3 support the GenServer to register a name on start via the :name option. Registered names are also automatically cleaned up on termination. The supported values are:

• an atom - the GenServer is registered locally (to the current node) with the given name using Process.register/2.

• {:global, term} - the GenServer is registered globally with the given term using the functions in the :global module.

• {:via, module, term} - the GenServer is registered with the given mechanism and name. The :via option expects a module that exports register_name/2, unregister_name/1, whereis_name/1 and send/2. One such example is the :global module which uses these functions for keeping the list of names of processes and their associated PIDs that are available globally for a network of Elixir nodes. Elixir also ships with a local, decentralized and scalable registry called Registry for locally storing names that are generated dynamically.

For example, we could start and register our Stack server locally as follows:

# Start the server and register it locally with name MyStack
{:ok, _} = GenServer.start_link(Stack, "hello", name: MyStack)

# Now messages can be sent directly to MyStack
GenServer.call(MyStack, :pop)
#=> "hello"

Once the server is started, the remaining functions in this module (call/3, cast/2, and friends) will also accept an atom, or any {:global, ...} or {:via, ...} tuples. In general, the following formats are supported:

• a PID
• an atom if the server is locally registered
• {atom, node} if the server is locally registered at another node
• {:global, term} if the server is globally registered
• {:via, module, name} if the server is registered through an alternative registry

If there is an interest to register dynamic names locally, do not use atoms, as atoms are never garbage-collected and therefore dynamically generated atoms won't be garbage-collected. For such cases, you can set up your own local registry by using the Registry module.

Receiving "regular" messages

The goal of a GenServer is to abstract the "receive" loop for developers, automatically handling system messages, supporting code change, synchronous calls and more. Therefore, you should never call your own "receive" inside the GenServer callbacks as doing so will cause the GenServer to misbehave.

Besides the synchronous and asynchronous communication provided by call/3 and cast/2, "regular" messages sent by functions such as send/2, Process.send_after/4 and similar, can be handled inside the handle_info/2 callback.

handle_info/2 can be used in many situations, such as handling monitor DOWN messages sent by Process.monitor/1. Another use case for handle_info/2 is to perform periodic work, with the help of Process.send_after/4:

defmodule MyApp.Periodically do
  use GenServer

  def start_link(_) do
    GenServer.start_link(__MODULE__, %{})
  end

  @impl true
  def init(state) do
    # Schedule work to be performed on start
    schedule_work()

    {:ok, state}
  end

  @impl true
  def handle_info(:work, state) do
    # Do the desired work here
    # ...

    # Reschedule once more
    schedule_work()

    {:noreply, state}
  end

  defp schedule_work do
    # We schedule the work to happen in 2 hours (written in milliseconds).
    # Alternatively, one might write :timer.hours(2)
    Process.send_after(self(), :work, 2 * 60 * 60 * 1000)
  end
end

Timeouts

The return value of init/1 or any of the handle_* callbacks may include a timeout value in milliseconds; if not, :infinity is assumed. The timeout can be used to detect a lull in incoming messages.

The timeout() value is used as follows:

• If the process has any message already waiting when the timeout() value is returned, the timeout is ignored and the waiting message is handled as usual. This means that even a timeout of 0 milliseconds is not guaranteed to execute (if you want to take another action immediately and unconditionally, use a :continue instruction instead).

• If any message arrives before the specified number of milliseconds elapse, the timeout is cleared and that message is handled as usual.

• Otherwise, when the specified number of milliseconds have elapsed with no message arriving, handle_info/2 is called with :timeout as the first argument.

When (not) to use a GenServer

So far, we have learned that a GenServer can be used as a supervised process that handles sync and async calls. It can also handle system messages, such as periodic messages and monitoring events. GenServer processes may also be named.

A GenServer, or a process in general, must be used to model runtime characteristics of your system. A GenServer must never be used for code organization purposes.

In Elixir, code organization is done by modules and functions, processes are not necessary. For example, imagine you are implementing a calculator and you decide to put all the calculator operations behind a GenServer:

def add(a, b) do
  GenServer.call(__MODULE__, {:add, a, b})
end

def subtract(a, b) do
  GenServer.call(__MODULE__, {:subtract, a, b})
end

def handle_call({:add, a, b}, _from, state) do
  {:reply, a + b, state}
end

def handle_call({:subtract, a, b}, _from, state) do
  {:reply, a - b, state}
end

This is an anti-pattern not only because it convolutes the calculator logic but also because you put the calculator logic behind a single process that will potentially become a bottleneck in your system, especially as the number of calls grow. Instead just define the functions directly:

def add(a, b) do
  a + b
end

def subtract(a, b) do
  a - b
end

If you don't need a process, then you don't need a process. Use processes only to model runtime properties, such as mutable state, concurrency and failures, never for code organization.

Debugging with the :sys module

GenServers, as special processes, can be debugged using the :sys module. Through various hooks, this module allows developers to introspect the state of the process and trace system events that happen during its execution, such as received messages, sent replies and state changes.

Let's explore the basic functions from the :sys module used for debugging:

• :sys.get_state/2 - allows retrieval of the state of the process. In the case of a GenServer process, it will be the callback module state, as passed into the callback functions as last argument.
• :sys.get_status/2 - allows retrieval of the status of the process. This status includes the process dictionary, if the process is running or is suspended, the parent PID, the debugger state, and the state of the behaviour module, which includes the callback module state (as returned by :sys.get_state/2). It's possible to change how this status is represented by defining the optional GenServer.format_status/1 callback.
• :sys.trace/3 - prints all the system events to :stdio.
• :sys.statistics/3 - manages collection of process statistics.
• :sys.no_debug/2 - turns off all debug handlers for the given process. It is very important to switch off debugging once we're done. Excessive debug handlers or those that should be turned off, but weren't, can seriously damage the performance of the system.
• :sys.suspend/2 - allows to suspend a process so that it only replies to system messages but no other messages. A suspended process can be reactivated via :sys.resume/2.

Let's see how we could use those functions for debugging the stack server we defined earlier.

iex> {:ok, pid} = Stack.start_link("")
iex> :sys.statistics(pid, true) # turn on collecting process statistics
iex> :sys.trace(pid, true) # turn on event printing
iex> Stack.push(pid, 1)
*DBG* <0.122.0> got cast {push,1}
*DBG* <0.122.0> new state [1]
:ok

iex> :sys.get_state(pid)
[1]

iex> Stack.pop(pid)
*DBG* <0.122.0> got call pop from <0.80.0>
*DBG* <0.122.0> sent 1 to <0.80.0>, new state []
1

iex> :sys.statistics(pid, :get)
{:ok,
 [
   start_time: {{2016, 7, 16}, {12, 29, 41}},
   current_time: {{2016, 7, 16}, {12, 29, 50}},
   reductions: 117,
   messages_in: 2,
   messages_out: 0
 ]}

iex> :sys.no_debug(pid) # turn off all debug handlers
:ok

iex> :sys.get_status(pid)
{:status, #PID<0.122.0>, {:module, :gen_server},
 [
   [
     "$initial_call": {Stack, :init, 1},            # process dictionary
     "$ancestors": [#PID<0.80.0>, #PID<0.51.0>]
   ],
   :running,                                        # :running | :suspended
   #PID<0.80.0>,                                    # parent
   [],                                              # debugger state
   [
     header: 'Status for generic server <0.122.0>', # module status
     data: [
       {'Status', :running},
       {'Parent', #PID<0.80.0>},
       {'Logged events', []}
     ],
     data: [{'State', [1]}]
   ]
 ]}

Learn more

If you wish to find out more about GenServers, the Elixir Getting Started guide provides a tutorial-like introduction. The documentation and links in Erlang can also provide extra insight.

• GenServer - Elixir's Getting Started Guide
• :gen_server module documentation
• gen_server Behaviour - OTP Design Principles
• Clients and Servers - Learn You Some Erlang for Great Good!