动态监督者与注册表
学习使用 `DynamicSupervisor` 动态启动和停止进程,并使用 `Registry` 查找进程。
动态监督者与注册表 是 CoddyKit 上的免费 Elixir & Phoenix: Scalable Backend Development 课时。 这是第 3 节课,共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 Elixir & Phoenix: Scalable Backend Development 学习路径的一部分,你的进度在网页和 CoddyKit 应用中同步。 Elixir & Phoenix: Scalable Backend Development 课程共包含 4 节课。
本课时的部分内容尚未翻译,以英文显示。
Dynamic Process Management
In previous lessons, we learned about supervisors for managing a fixed set of processes. But what if your application needs to create or destroy processes on demand, like for each user session or a specific task?
This is where dynamic process management comes in handy. It allows your application to adapt and scale by creating processes only when they are needed.
Introducing DynamicSupervisor
Elixir's DynamicSupervisor is designed precisely for this purpose. Unlike Supervisor, which manages a predefined list of children, a DynamicSupervisor starts with no children.
It provides functions to dynamically add and remove child processes during runtime, giving you flexible control over your application's process tree.
Starting a DynamicSupervisor
To use a DynamicSupervisor, you first need to start it, typically within your application's supervision tree. Here's a basic setup:
defmodule MyApp.DynamicManager do
use DynamicSupervisor
def start_link(init_arg) do
DynamicSupervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
end
@impl true
def init(_init_arg) do
DynamicSupervisor.init(strategy: :one_for_one)
end
end
# To run this in an IEx session:
# {:ok, pid} = MyApp.DynamicManager.start_link([])
# IO.puts("DynamicSupervisor started with PID: #{inspect(pid)}")Adding Children Dynamically
Once your DynamicSupervisor is running, you can add new children using DynamicSupervisor.start_child/2. Each child needs a child specification defining how it should be started.
Let's create a simple GenServer that the supervisor can manage.
defmodule MyApp.Worker do
use GenServer
def start_link(id) do
GenServer.start_link(__MODULE__, id, name: via_tuple(id))
end
def init(id) do
IO.puts("Worker #{id} started!")
{:ok, id}
end
defp via_tuple(id), do: {:via, Registry, {MyApp.Registry, id}}
end
defmodule MyApp.DynamicManager do
use DynamicSupervisor
def start_link(init_arg) do
DynamicSupervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
end
@impl true
def init(_init_arg) do
DynamicSupervisor.init(strategy: :one_for_one)
end
def start_worker(id) do
child_spec = %{
id: id,
start: {MyApp.Worker, :start_link, [id]}
}
DynamicSupervisor.start_child(__MODULE__, child_spec)
end
end
# To run this in an IEx session:
# {:ok, manager_pid} = MyApp.DynamicManager.start_link([])
# {:ok, worker_pid} = MyApp.DynamicManager.start_worker(:user_123)
# IO.puts("Worker PID: #{inspect(worker_pid)}")Stopping Dynamic Children
Just as you can start children dynamically, you can also stop them. Use DynamicSupervisor.terminate_child/2 with the child's PID.
The supervisor will handle the graceful shutdown of the process.
defmodule MyApp.Worker do
use GenServer
def start_link(id) do
GenServer.start_link(__MODULE__, id, name: {:global, id})
end
def init(id) do
IO.puts("Worker #{id} started!")
{:ok, id}
end
def terminate(_reason, id) do
IO.puts("Worker #{id} stopping!")
end
end
defmodule MyApp.DynamicManager do
use DynamicSupervisor
def start_link(init_arg) do
DynamicSupervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
end
@impl true
def init(_init_arg) do
DynamicSupervisor.init(strategy: :one_for_one)
end
def start_worker(id) do
child_spec = %{id: id, start: {MyApp.Worker, :start_link, [id]}}
DynamicSupervisor.start_child(__MODULE__, child_spec)
end
def stop_worker(worker_pid) do
DynamicSupervisor.terminate_child(__MODULE__, worker_pid)
end
end
# To run this in an IEx session:
# {:ok, manager_pid} = MyApp.DynamicManager.start_link([])
# {:ok, worker_pid} = MyApp.DynamicManager.start_worker(:task_A)
# :timer.sleep(100)
# MyApp.DynamicManager.stop_worker(worker_pid)
# :timer.sleep(100)
# IO.puts("Worker process should now be stopped.")The Need for a Registry
When you start processes dynamically, how do you find them later? If you only have their PIDs, what happens if a process crashes and restarts with a new PID?
We need a way to assign a persistent, meaningful name to a process and look it up reliably, regardless of its current PID. This is where Elixir's Registry comes in.
What is Registry?
The Registry module provides a local, distributed, and fault-tolerant key-value store for associating arbitrary terms (keys) with process identifiers (PIDs).
It's perfect for looking up processes by a name or ID, especially when dealing with dynamic processes or when you want to use a custom name instead of global or local atoms.
Starting and Using Registry
Like other OTP behaviors, Registry needs to be started, typically as part of your application's supervision tree. You can define multiple registries, each with its own scope.
Once started, you can register processes with keys and look them up.
defmodule MyApp.Registry do
use Registry, keys: :unique, name: __MODULE__
end
defmodule MyApp.WorkerWithRegistry do
use GenServer
def start_link(id) do
GenServer.start_link(__MODULE__, id, name: {:via, Registry, {MyApp.Registry, id}})
end
def init(id) do
IO.puts("Worker #{id} started and registered!")
{:ok, id}
end
def get_state(pid) do
GenServer.call(pid, :get_state)
end
@impl true
def handle_call(:get_state, _from, state) do
{:reply, state, state}
end
end
# To run this in an IEx session:
# {:ok, _} = MyApp.Registry.start_link([]) # Start the registry
# {:ok, worker_pid} = MyApp.WorkerWithRegistry.start_link(:session_1)
# IO.puts("Registered worker PID: #{inspect(worker_pid)}")
#
# {:ok, [pid_found]} = Registry.lookup(MyApp.Registry, :session_1)
# IO.puts("Looked up PID: #{inspect(pid_found)}")
# IO.puts("Worker state: #{MyApp.WorkerWithRegistry.get_state(pid_found)}")DynamicSupervisor with Registry
Combining DynamicSupervisor with Registry is a powerful pattern. You can dynamically start worker processes, and each worker registers itself with a unique key in the Registry.
This allows other parts of your application to find and interact with specific worker processes without needing their PIDs directly.
defmodule MyApp.UserSession do
use GenServer
def start_link(user_id) do
# Register with Registry using the user_id as key
GenServer.start_link(__MODULE__, user_id, name: {:via, Registry, {MyApp.SessionsRegistry, user_id}})
end
def init(user_id) do
IO.puts("Session for user #{user_id} started.")
{:ok, user_id}
end
def get_user_id(pid) do
GenServer.call(pid, :get_user_id)
end
@impl true
def handle_call(:get_user_id, _from, user_id) do
{:reply, user_id, user_id}
end
end
defmodule MyApp.SessionsRegistry do
use Registry, keys: :unique, name: __MODULE__
end
defmodule MyApp.SessionSupervisor do
use DynamicSupervisor
def start_link(init_arg) do
DynamicSupervisor.start_link(__MODULE__, init_arg, name: __MODULE__)
end
@impl true
def init(_init_arg) do
DynamicSupervisor.init(strategy: :one_for_one)
end
def start_session(user_id) do
child_spec = %{
id: user_id,
start: {MyApp.UserSession, :start_link, [user_id]}
}
DynamicSupervisor.start_child(__MODULE__, child_spec)
end
end
# --- Application Entry Point --- (Simulated)
{:ok, _registry_pid} = MyApp.SessionsRegistry.start_link([])
{:ok, _supervisor_pid} = MyApp.SessionSupervisor.start_link([])
# Start a session for user_101
{:ok, session_pid_1} = MyApp.SessionSupervisor.start_session(:user_101)
IO.puts("Session 1 PID: #{inspect(session_pid_1)}")
# Start a session for user_102
{:ok, session_pid_2} = MyApp.SessionSupervisor.start_session(:user_102)
IO.puts("Session 2 PID: #{inspect(session_pid_2)}")
# Look up user_101's session via Registry
{:ok, [pid_from_registry]} = Registry.lookup(MyApp.SessionsRegistry, :user_101)
IO.puts("Found user_101 session PID via Registry: #{inspect(pid_from_registry)}")
IO.puts("User ID from session: #{MyApp.UserSession.get_user_id(pid_from_registry)}")Dynamic & Registry Quiz
Consider an application that uses DynamicSupervisor to manage user sessions and Registry to look up sessions by user ID. What is the primary benefit of using Registry in this scenario?
Recap: Dynamic & Registry
You've mastered dynamic process management!
DynamicSupervisorallows you to start and stop child processes on demand, making your application highly flexible and adaptable.- The
Registrymodule provides a powerful key-value store for associating processes with meaningful names or IDs, enabling reliable lookup of dynamically started processes.
By combining these two, you can build robust, scalable applications that manage their resources efficiently.
用 AI 导师学习 Elixir — 免费
在浏览器中编写并运行真实代码,获得全天候 AI 导师的即时帮助,并在网页或应用中继续学习。
- 课程
- 12
- 课程
- 48
常见问题解答
「动态监督者与注册表」课时是免费的吗?
是的 — 「动态监督者与注册表」的完整文本可在网页上免费阅读。要进行交互式练习(内置代码编辑器和全天候 AI 导师)并解锁 Elixir & Phoenix: Scalable Backend Development 课程的其余内容,请升级到 CoddyKit PRO。 Elixir & Phoenix: Scalable Backend Development 课程共包含 4 节课。
「动态监督者与注册表」这节课中我会学到什么?
学习使用 `DynamicSupervisor` 动态启动和停止进程,并使用 `Registry` 查找进程。 你通过在浏览器中直接运行的动手代码来练习 Elixir & Phoenix: Scalable Backend Development,全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 Elixir & Phoenix: Scalable Backend Development 需要有经验吗?
无需任何先前经验。CoddyKit 上的 Elixir & Phoenix: Scalable Backend Development 课程适合初学者到高级学习者,你可以从这里开始或从头开始,按照自己的节奏学习。 这是第 3 节课,共 4 节。
「动态监督者与注册表」课时需要多长时间?
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动,所以你能稳步进步,并在网页和应用中从离开的地方继续。
我能在这节 Elixir & Phoenix: Scalable Backend Development 课中编写并运行代码吗?
能。每节 Elixir & Phoenix: Scalable Backend Development 课都包含内置代码编辑器,你可以在浏览器中直接编写并运行真实代码,并获得即时 AI 反馈 — 无需本地设置。
此课程中的所有课时
- 分布式 Elixir 与集群
- 高级监督策略
- 动态监督者与注册表
- GenStage 与背压流水线