Helm: The Package Manager for Kubernetes

Why it exists

Helm was created to solve the problem of managing applications in Kubernetes. Without a package manager, deploying and upgrading complex applications involves handling numerous individual resource files, which is error-prone. Helm provides a standardized way to bundle these resources and manage their lifecycle as a single unit.

The mental model

Think of Helm as the package manager for Kubernetes, directly analogous to tools like apt for Debian or yum for Red Hat. Just as you use a system package manager to install, upgrade, and remove software on a server, you use Helm to do the same for applications on a Kubernetes cluster. Its status as a Cloud Native Computing Foundation (CNCF) graduated project signifies its maturity and wide adoption.

How it works

Helm's functionality is centered around a packaging format called Charts. A Chart is a collection of files that describe a related set of Kubernetes resources. The documentation mentions a "Chart Template Guide," which implies that Charts use templates to generate manifest files, allowing for customization during installation. Helm also provides a Go SDK for programmatic interaction and a command-line interface with dozens of commands.

When to use it

Use Helm whenever you need to deploy applications to Kubernetes. This applies to both installing third-party software and packaging your own applications. The project maintains a searchable repository of over 800 charts for common applications, and provides extensive documentation on "Developing Charts" for your own software, covering topics from best practices to plugins.

When not to use it

The official documentation does not specify scenarios where Helm is inappropriate. Its design and position as a CNCF graduated project suggest it is intended as a general-purpose solution for the majority of Kubernetes application management use cases.

One canonical example

A canonical example is using Helm to install a piece of common software, like a database, from the public chart repository. While the source does not provide a specific command, it references a searchable repository of "800+ Charts," implying a primary workflow is to find a pre-existing chart and deploy it, rather than building everything from scratch. Another example is developing your own chart for your application, following the official best practices.