Book Image

Feature Management with LaunchDarkly

By : Michael Gillett
Book Image

Feature Management with LaunchDarkly

By: Michael Gillett

Overview of this book

Over the past few years, DevOps has become the de facto approach for designing, building, and delivering software. Feature management is now extending the DevOps methodology to allow applications to change on demand and run experiments to validate the success of new features. If you want to make feature management happen, LaunchDarkly is the tool for you. This book explains how feature management is key to building modern software systems. Starting with the basics of LaunchDarkly and configuring simple feature flags to turn features on and off, you'll learn how simple functionality can be applied in more powerful ways with percentage-based rollouts, experimentation, and switches. You'll see how feature management can change the way teams work and how large projects, including migrations, are planned. Finally, you'll discover various uses of every part of the tool to gain mastery of LaunchDarkly. This includes tips and tricks for experimentation, identifying groups and segments of users, and investigating and debugging issues with specific users and feature flag evaluations. By the end of the book, you'll have gained a comprehensive understanding of LaunchDarkly, along with knowledge of the adoption of trunk-based development workflows and methods, multi-variant testing, and managing infrastructure changes and migrations.
Table of Contents (18 chapters)
1
Section 1: The Basics
5
Section 2:Getting the Most out of Feature Management
11
Section 3: Mastering LaunchDarkly

What is an experiment?

So far in this book, I have tried to avoid the words test or experiment, but ultimately, a large part of what we've explored so far has been about running tests and experimenting with new code. When we looked at rollouts in Chapter 4, Percentage and Ring Rollouts, I emphasized validating whether the features are technically working and that they are adding the expected value to customers. This can be termed as an experiment.

While I did not state the validation of a rollout in the terms of an experiment, there was effectively a hypothesis being validated by rolling out a feature to a percentage or ring of the customers. For example, when ensuring that something is technically working as expected, the hypothesis would be that the new code is going to work as expected and meet the technical requirements. When we enable that for 10% of the customers and gather telemetry to ensure that this is indeed the case, we are testing the hypothesis. By rolling that...