Elixir & Phoenix: Scalable Backend Development · レッスン

Telemetryとメトリクスによる監視

アプリケーションにTelemetryを統合してメトリクスを出力し、実行時の動作を把握します。

「Telemetryとメトリクスによる監視」はCoddyKit上の無料Elixir & Phoenix: Scalable Backend Developmentレッスンです。これはレッスン2/4です。下記で完全なレッスンを無料で読むことができます。その後、ブラウザ内の組み込みコードエディタと24時間対応のAIチューターでハンズオン演習できます。これはElixir & Phoenix: Scalable Backend Development学習パスの一部であり、ウェブとCoddyKitアプリ全体で進捗が同期されます。 Elixir & Phoenix: Scalable Backend Developmentコースには全4レッスンが含まれています。

このレッスンの一部はまだ翻訳されておらず、英語で表示されています。

What is Elixir Telemetry?

In Elixir, Telemetry is a powerful, low-level event dispatching library. It helps you understand what's happening inside your running application.

It's built into Elixir/OTP, making it a standard way to observe systems.
Think of it as a central nervous system for your app's events.
It enables you to collect metrics, trace operations, and monitor performance.

It's crucial for building observable and maintainable applications.

Events: The Core of Telemetry

Telemetry works by emitting events. An event is a signal that something notable has happened in your application.

Each event has a unique name (a list of atoms, e.g., [:my_app, :user, :login]).
Events can carry measurements (numeric values like duration) and metadata (additional context like user ID).
You don't emit events directly; instead, you use functions that wrap your code.

Measuring Duration with Telemetry.span

One common use of Telemetry is to measure how long an operation takes. This is done using :telemetry.span/3.

It executes a given function and measures its execution time.
It emits two events: one when the span starts and one when it ends.
The end event includes the duration as a measurement.

This is great for profiling function calls or database queries.

Emitting Discrete Events with Telemetry.execute

Sometimes, you just want to signal that an action occurred, without necessarily measuring its duration. For this, use :telemetry.execute/3.

It emits a single event with specified measurements and metadata.
Useful for tracking things like cache hits/misses, background job completions, or specific API calls.
It's a simpler, more direct way to emit a single point-in-time event.

Code: Emitting a Telemetry Span

Let's see :telemetry.span/3 in action. This example simulates a 'processing' task and measures its duration.

We define a simple module and call a function that wraps its work in a span.

defmodule MyApp.Worker do
  require Logger

  def process_data(data) do
    Logger.info("Starting data processing...")
    :telemetry.span(
      [:my_app, :worker, :process_data],
      %{input_length: byte_size(data)},
      fn ->
        # Simulate some work
        :timer.sleep(100)
        result = String.upcase(data)
        {:ok, result}
      end
    )
  end
end

# --- Main execution for CoddyKit ---
Logger.info("Calling MyApp.Worker.process_data...")
{:ok, result} = MyApp.Worker.process_data("hello elixir")
IO.puts("Processed result: #{result}")
IO.puts("\n(No Telemetry output yet - we need a handler!)")

Attaching Telemetry Handlers

Emitting events is only half the story! To do something useful with them, you need to attach handlers.

A handler is a function that gets called whenever a specific Telemetry event is emitted.
You attach a handler using :telemetry.attach/4, specifying the event name and your handler function.
Handlers allow you to react to events: log them, send them to a metrics system, or trigger other actions.

They are the listeners of your application's internal signals.

Handler Function Structure

A Telemetry handler function must accept four arguments:

event_name: The name of the event (e.g., [:my_app, :user, :login]).
measurements: A map of numeric values associated with the event (e.g., %{duration: 123456}).
metadata: A map of additional context (e.g., %{user_id: 123}).
config: Any custom configuration passed during attachment.

Your handler logic will use these arguments to process the event.

Code: Attaching and Handling Events

Now let's add a handler to our previous example. This handler will simply log the event details.

Notice how :telemetry.attach/4 links our handler function to the specific event name.

defmodule MyApp.WorkerWithHandler do
  require Logger

  # Define our simple handler function
  def handle_event(event_name, measurements, metadata, _config) do
    Logger.info("\n--- Telemetry Event Received ---")
    Logger.info("Event: #{inspect(event_name)}")
    Logger.info("Measurements: #{inspect(measurements)}")
    Logger.info("Metadata: #{inspect(metadata)}")
    Logger.info("------------------------------")
  end

  # Function that emits a Telemetry event
  def process_data(data) do
    Logger.info("Starting data processing...")
    :telemetry.span(
      [:my_app, :worker, :process_data],
      %{input_length: byte_size(data)},
      fn ->
        :timer.sleep(100)
        result = String.upcase(data)
        {:ok, result}
      end
    )
  end
end

# --- Main execution for CoddyKit ---
# Attach the handler *before* emitting events
:telemetry.attach(
  "my-worker-handler",
  [:my_app, :worker, :process_data], # Event name to listen for
  &MyApp.WorkerWithHandler.handle_event/4, # Our handler function
  nil # Optional config
)
Logger.info("Telemetry handler 'my-worker-handler' attached.")

# Perform some work, which will now trigger the handler
{:ok, result} = MyApp.WorkerWithHandler.process_data("hello world")
IO.puts("Processed result: #{result}")

# Clean up: detach the handler (good practice in tests/scripts)
:telemetry.detach("my-worker-handler")
Logger.info("Telemetry handler 'my-worker-handler' detached.")

Telemetry as a Metrics Source

Telemetry is not a metrics system itself, but it's an excellent source for them. By attaching handlers, you can feed events into dedicated metrics libraries.

Libraries like telemetry_metrics can aggregate Telemetry events into counters, gauges, and histograms.
These aggregated metrics are then often exported to monitoring systems like Prometheus or Datadog.
This separation keeps Telemetry lightweight and flexible, allowing you to choose your preferred metrics backend.

Quick Check: Telemetry Handlers

You've learned about emitting and handling Telemetry events. Let's test your understanding.

Recap: Telemetry for Observability

In this lesson, we explored Elixir's Telemetry library:

Telemetry is an event dispatching system for internal application observability.
It uses events with unique names, measurements, and metadata.
:telemetry.span/3 measures code execution duration.
:telemetry.execute/3 emits discrete, point-in-time events.
Handlers are functions attached via :telemetry.attach/4 to process events.
Telemetry events are a prime source for building application metrics.

Mastering Telemetry is key to understanding and monitoring your Elixir applications effectively!

よくある質問

「Telemetryとメトリクスによる監視」レッスンは無料ですか？

はい。「Telemetryとメトリクスによる監視」の完全なテキストはこのウェブで無料で読めます。インタラクティブに演習し（組み込みコードエディタと24時間対応のAIチューター）、Elixir & Phoenix: Scalable Backend Developmentコースの残りをアンロックするには、CoddyKit PROにアップグレードしてください。 Elixir & Phoenix: Scalable Backend Developmentコースには全4レッスンが含まれています。

「Telemetryとメトリクスによる監視」で何を学びますか？

アプリケーションにTelemetryを統合してメトリクスを出力し、実行時の動作を把握します。ブラウザで直接実行するハンズオンコードでElixir & Phoenix: Scalable Backend Developmentを演習し、24時間対応のAIチューターがレッスンを進める中での質問に答えます。

Elixir & Phoenix: Scalable Backend Developmentを始めるのに経験は必要ですか？

事前経験は必要ありません。CoddyKitのElixir & Phoenix: Scalable Backend Developmentは初級者から上級者向けに構成されているため、ここから始めるか最初から始めて、自分のペースで進むことができます。これはレッスン2/4です。

「Telemetryとメトリクスによる監視」レッスンにはどのくらい時間がかかりますか？

ほとんどのCoddyKitレッスンは約5～10分かかります。各レッスンはコンパクトでインタラクティブなので、着実に進歩し、ウェブとアプリ全体で正確に前回の場所から再開できます。

このElixir & Phoenix: Scalable Backend Developmentレッスンでコードを書いて実行できますか？

はい。すべてのElixir & Phoenix: Scalable Backend Developmentレッスンに組み込みコードエディタが含まれているため、ブラウザでリアルコードを書いて実行し、即座のAIフィードバックを取得できます。ローカル設定は不要です。

このコースのすべてのレッスン

← Elixir & Phoenix: Scalable Backend Developmentに戻る