Load Testing & Performance Benchmarking (JMeter & k6) · レッスン

k6メトリクスの読み解き方

k6の詳細なメトリクスとカスタムタグを理解し、分析に役立つ形で可視化する方法を学びます。

「k6メトリクスの読み解き方」はCoddyKit上の無料Load Testing & Performance Benchmarking (JMeter & k6)レッスンです。これはレッスン3/4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Load Testing & Performance Benchmarking (JMeter & k6)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。 Load Testing & Performance Benchmarking (JMeter & k6)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 are k6 Metrics?

When you run a k6 performance test, it generates various metrics. These are numerical values that describe your system's behavior under load.

Understanding these metrics is crucial for identifying performance bottlenecks, verifying service level agreements (SLAs), and ensuring your application can handle expected traffic.

Key Metric: Request Duration

One of the most important built-in k6 metrics is http_req_duration. This measures the total time it takes for an HTTP request to complete.

It starts when the request is sent.
It ends when the full response is received.
It directly reflects the response time users experience.

Key Metrics: VUs & Iterations

Other fundamental metrics help you understand the load:

vus (Virtual Users): Shows the number of virtual users currently active during the test.
iterations: Counts how many times your test script's default function has been executed.

These metrics help you see if your test is applying the intended load.

Key Metrics: Data Transfer

Understanding network usage is also vital. k6 provides metrics for data transfer:

data_sent: The total number of bytes sent by k6 during the test.
data_received: The total number of bytes received by k6.

These metrics help analyze bandwidth consumption and identify potentially large responses.

Understanding Metric Aggregations

k6 doesn't just give raw values; it aggregates them. Common aggregations include:

avg (average)
min (minimum)
max (maximum)
p(XX) (percentile, e.g., p(90), p(95))

Percentiles are especially important as they represent the experience of a certain percentage of your users, ignoring extreme outliers.

Beyond Built-in: Custom Metrics

While k6's built-in metrics are powerful, you might need to track specific application-level events or durations. This is where custom metrics come in.

You can create your own metrics to gain deeper insights into your system's unique behavior. Common types are Counter, Gauge, Rate, and Trend.

Adding a Custom Metric

Here's how to add a Trend custom metric to measure the duration of a specific operation within your script. This helps track performance beyond just HTTP requests.

import { Trend } from 'k6/metrics';
import http from 'k6/http';
import { sleep } from 'k6';

const myCustomTrend = new Trend('custom_operation_duration');

export default function () {
  const startTime = Date.now();
  http.get('https://test.k6.io/contacts.php');
  const endTime = Date.now();
  myCustomTrend.add(endTime - startTime); // Add duration in milliseconds

  sleep(1);
}

Categorizing with k6 Tags

Tags are key-value pairs that provide metadata for your metrics. They allow you to categorize and filter your results for more granular analysis.

k6 automatically adds tags like url and status. You can also add custom tags to differentiate metrics for different parts of your application or specific user flows.

Applying Custom Tags

Custom tags are very useful for breaking down metrics. For example, you can tag requests based on the page or API endpoint they hit, allowing you to analyze performance per feature.

import http from 'k6/http';
import { sleep } from 'k6';

export default function () {
  // Tagging requests to different pages
  http.get('https://test.k6.io/', {
    tags: { page_name: 'homepage' },
  });

  http.get('https://test.k6.io/news.php', {
    tags: { page_name: 'news_page' },
  });

  sleep(1);
}

Visualizing Your k6 Metrics

While k6's console output is useful, visualizing your metrics makes analysis much easier. Tools like Grafana (often paired with InfluxDB or Prometheus) are excellent for this.

k6 can export results directly to these systems, allowing you to create dashboards that show trends, spikes, and performance regressions over time.

Metric Interpretation Challenge

Consider the k6 output snippet below:

http_req_duration{status=200, scenario=default, my_tag=checkout}......: avg=350ms min=100ms med=300ms p(90)=550ms p(95)=600ms max=1200ms

What does p(95)=600ms primarily tell you about the checkout requests?

Recap: Interpreting k6 Metrics

You've learned how to interpret k6's powerful metrics!

Built-in metrics like http_req_duration, vus, and data_sent/received give foundational insights.
Aggregations, especially percentiles, provide a realistic view of user experience.
Custom metrics and tags allow for deep, granular analysis of specific application behaviors.
Visualization tools are essential for making sense of complex test results.

Mastering metric interpretation is key to successful performance testing!

よくある質問

「k6メトリクスの読み解き方」レッスンは無料ですか？

はい。「k6メトリクスの読み解き方」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）、Load Testing & Performance Benchmarking (JMeter & k6)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。 Load Testing & Performance Benchmarking (JMeter & k6)コースには全4レッスンが含まれています。

「k6メトリクスの読み解き方」で何を学びますか？

k6の詳細なメトリクスとカスタムタグを理解し、分析に役立つ形で可視化する方法を学びます。ブラウザで直接実行するハンズオンコードでLoad Testing & Performance Benchmarking (JMeter & k6)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。

Load Testing & Performance Benchmarking (JMeter & k6)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Load Testing & Performance Benchmarking (JMeter & k6)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3/4です。

「k6メトリクスの読み解き方」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Load Testing & Performance Benchmarking (JMeter & k6)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Load Testing & Performance Benchmarking (JMeter & k6)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

← Load Testing & Performance Benchmarking (JMeter & k6)に戻る