Supervisors and Application Structure
Design resilient applications using supervisors to monitor and restart processes, ensuring fault tolerance.
Supervisors and Application Structure is a free Elixir & Phoenix: Scalable Backend Development lesson on CoddyKit — lesson 3 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the Elixir & Phoenix: Scalable Backend Development learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
Building Resilient Systems
In concurrent applications, things can go wrong. Processes might crash due to unexpected errors or external issues.
Fault tolerance is the ability of a system to continue operating even when components fail. Elixir embraces a "let it crash" philosophy, meaning instead of trying to prevent every single crash, it focuses on gracefully recovering from them.
Guardians of Processes
This is where Supervisors come in! A supervisor is a special kind of process designed to monitor other processes (its "children").
- If a child process crashes, the supervisor automatically restarts it.
- This ensures your application remains stable and available.
- Supervisors form the backbone of fault-tolerant Elixir applications.
Your First Supervisor
Let's define a simple supervisor module. It uses the Supervisor behavior, similar to how GenServer uses the GenServer behavior.
The init/1 callback is where you define the children it will supervise and its restart strategy.
defmodule MySupervisor do
use Supervisor
def start_link(init_arg) do
Supervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
end
@impl true
def init(_init_arg) do
# No children yet, just the supervisor itself
children = []
Supervisor.init(children, strategy: :one_for_one)
end
end
# --- Main execution part ---
# This simulates starting the supervisor and checking its status.
# In a real app, this would be part of an Application's start/2.
IO.puts("Attempting to start MySupervisor...")
{:ok, supervisor_pid} = MySupervisor.start_link([])
IO.puts("MySupervisor started with PID: #{inspect(supervisor_pid)}")
# Check if the supervisor process is alive
if Process.alive?(supervisor_pid) do
IO.puts("Supervisor is alive!")
else
IO.puts("Supervisor is NOT alive!")
endUnderstanding Supervision Strategies
Supervisors have different strategies for handling child failures:
:one_for_one: Restarts only the child that crashed. This is the default and most common.:one_for_all: If any child crashes, all other children are terminated and then all children are restarted.:rest_for_one: If a child crashes, it and all children started *after* it are terminated and then restarted.
Choosing the right strategy depends on the dependencies between your processes.
Adding Supervised Children
To make a supervisor useful, it needs children! You define children using Supervisor.child_spec/2, which tells the supervisor how to start and manage a process.
Here, we define a simple MyWorker GenServer and add it as a child to our supervisor.
defmodule MyWorker do
use GenServer
def start_link(_opts) do
GenServer.start_link(__MODULE__, :ok, name: __MODULE__)
end
@impl true
def init(:ok) do
IO.puts("MyWorker started!")
{:ok, %{}}
end
@impl true
def handle_call(:crash, _from, state) do
IO.puts("MyWorker is crashing!")
exit(:bad_state) # Simulate a crash
{:reply, :ok, state} # This line won't be reached
end
end
defmodule MySupervisorWithWorker do
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 = [
# Define our worker as a child process
Supervisor.child_spec(MyWorker, id: MyWorker)
]
Supervisor.init(children, strategy: :one_for_one)
end
end
# --- Main execution part ---
IO.puts("Starting supervisor with worker...")
{:ok, supervisor_pid} = MySupervisorWithWorker.start_link([])
IO.puts("Supervisor PID: #{inspect(supervisor_pid)}")
# Get the worker's PID
worker_pid = Process.whereis(MyWorker)
IO.puts("Initial MyWorker PID: #{inspect(worker_pid)}")
if Process.alive?(worker_pid) do
IO.puts("Worker is alive and supervised.")
else
IO.puts("Worker did not start correctly.")
endFault Tolerance in Action
Now, let's see the supervisor in action! We'll deliberately crash our MyWorker process, and the supervisor will automatically restart it.
Notice how the worker's Process ID (PID) changes, indicating a new process was spawned.
defmodule MyWorker do
use GenServer
def start_link(_opts) do
GenServer.start_link(__MODULE__, :ok, name: __MODULE__)
end
@impl true
def init(:ok) do
IO.puts("MyWorker started!")
{:ok, %{}}
end
@impl true
def handle_call(:crash, _from, state) do
IO.puts("MyWorker is crashing!")
exit(:bad_state) # Simulate a crash
{:reply, :ok, state} # This line won't be reached
end
def crash_it do
GenServer.call(__MODULE__, :crash)
end
end
defmodule MySupervisorWithWorker do
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 = [
Supervisor.child_spec(MyWorker, id: MyWorker)
]
Supervisor.init(children, strategy: :one_for_one)
end
end
# --- Main execution part ---
IO.puts("Starting supervisor with worker...")
{:ok, _supervisor_pid} = MySupervisorWithWorker.start_link([])
worker_pid_before_crash = Process.whereis(MyWorker)
IO.puts("Worker PID before crash: #{inspect(worker_pid_before_crash)}")
# Crash the worker
IO.puts("Attempting to crash the worker...")
MyWorker.crash_it()
:timer.sleep(100) # Give supervisor a moment to restart
worker_pid_after_crash = Process.whereis(MyWorker)
IO.puts("Worker PID after crash: #{inspect(worker_pid_after_crash)}")
if worker_pid_before_crash != worker_pid_after_crash && Process.alive?(worker_pid_after_crash) do
IO.puts("Worker was restarted by the supervisor! New PID detected.")
else
IO.puts("Worker was NOT restarted, or PID remained the same (unexpected).")
endElixir Applications: The Top Level
While supervisors manage individual processes, an Elixir Application is the top-level unit of code and processes in an Elixir system.
It provides a structured way to:
- Group related modules and processes.
- Define how your system starts up and shuts down.
- Manage configuration and dependencies.
The `Application` Behavior
Every Elixir application typically has a main application module that use Application.
The most important callback is start/2, which is invoked when your application starts. This is where you typically start your top-level supervisor, which then recursively starts all other processes in your system.
defmodule MyApp.Application do
use Application
# This is the entry point for your application.
# It starts the top-level supervisor.
@impl true
def start(_type, _args) do
children = [
# In a real app, you'd start your main supervisor here.
# For example: Supervisor.child_spec(MySupervisorWithWorker, id: MySupervisorWithWorker)
]
# Start a supervisor that will supervise other processes/supervisors
opts = [strategy: :one_for_one, name: MyApp.Supervisor]
Supervisor.start_link(children, opts)
end
end
# --- Main execution part ---
# This part simulates how an application would be started.
# In a real Mix project, 'mix run --no-halt' would call MyApp.Application.start/2
IO.puts("Simulating application start...")
{:ok, pid} = MyApp.Application.start(:normal, [])
IO.puts("Application top-level supervisor started with PID: #{inspect(pid)}")
if Process.alive?(pid) do
IO.puts("Application supervisor is active.")
else
IO.puts("Application supervisor failed to start.")
endHierarchical Supervision Trees
For complex applications, you often don't have just one supervisor. You create a supervision tree, where supervisors can supervise other supervisors.
- This allows you to organize your application into logical units.
- Different parts of your system can have different restart strategies.
- If a major component fails, its supervisor can restart it without affecting unrelated parts of the system.
Supervisor Check
Time for a quick check on your understanding of supervisor strategies!
Lesson Summary
Well done! You've learned how Elixir builds resilient applications:
- Supervisors monitor processes and restart them upon failure.
- Different supervision strategies (
:one_for_one,:one_for_all,:rest_for_one) dictate how failures are handled. - Elixir Applications provide the top-level structure, starting supervisors and forming supervision trees.
These concepts are fundamental to building robust, fault-tolerant systems in Elixir.
Frequently asked questions
Is the “Supervisors and Application Structure” lesson free?
Yes — the full text of “Supervisors and Application Structure” is free to read here on the web, and the Elixir & Phoenix: Scalable Backend Development course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the Elixir & Phoenix: Scalable Backend Development course, upgrade to CoddyKit PRO.
What will I learn in “Supervisors and Application Structure”?
Design resilient applications using supervisors to monitor and restart processes, ensuring fault tolerance. You practise Elixir & Phoenix: Scalable Backend Development with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start Elixir & Phoenix: Scalable Backend Development?
No prior experience is required. Elixir & Phoenix: Scalable Backend Development on CoddyKit is structured for beginners through advanced learners; this is — lesson 3 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Supervisors and Application Structure” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this Elixir & Phoenix: Scalable Backend Development lesson?
Yes. Every Elixir & Phoenix: Scalable Backend Development lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- Elixir Processes and Message Passing
- Implementing GenServer Behavior
- Supervisors and Application Structure
- Concurrent Work with Task and Agent