Pular para o conteúdo principal
Versão: 3.21.1

Introduction to Helm

Helm is the package manager for Kubernetes. It helps you define, install, and upgrade applications on a Kubernetes cluster, from a single container to an application with many interdependent parts.

Deploying an application to Kubernetes means writing and maintaining many manifests: Deployments, Services, ConfigMaps, and more. Managing those files by hand across environments and versions is repetitive and error prone. Helm packages these related manifests into a single unit called a chart, which you can version, share, install, and roll back as one release. This lets you treat an application the way a system package manager such as Homebrew, apt, or yum treats software on an operating system.

What can Helm do?​

You can use Helm to:

  • Install and manage applications. Deploy off-the-shelf applications, such as databases, monitoring stacks, and ingress controllers, from a chart instead of assembling manifests yourself.
  • Package and share your own applications. Bundle your Kubernetes resources into a chart, version it, and distribute it to your team or the wider community.
  • Configure applications per environment. Supply different values to the same chart to deploy to development, staging, and production without duplicating manifests.
  • Upgrade and roll back safely. Move a release to a new version, and return to a previous revision when an upgrade does not go as planned.
  • Manage dependencies. Declare the other charts your application needs, and let Helm install them together.

Who is Helm For?​

A Helm user often performs one of several roles. One person can perform more than one of these roles, and how the roles map to people varies between organizations. For more information, see User Profiles in the Helm Community documentation.

  • Application operator. You take an application and run it inside a Kubernetes cluster, such as operating WordPress and its MySQL database. This role differs from a cluster operator, who runs the cluster itself.
  • Application distributor. You package an application so that someone else can operate it, as the maintainers of community charts do.
  • Application developer. You write the software for an application, and are typically not concerned with where it runs.
  • Supporting tool developer. You build tools that work alongside Helm, such as a linter or a Helm plugin.

Helm focuses on the application running in the cluster rather than the cluster itself. Standing up and operating a Kubernetes cluster, including its control plane and nodes, is the work of a cluster operator and falls outside Helm's scope.

Key Components​

Three components describe how Helm works: charts, repositories, and releases. Helm installs charts into Kubernetes, creating a new release for each installation, and you find new charts by searching Helm chart repositories.

Chart​

A chart is a Helm package. It contains all of the resource definitions needed to run an application, tool, or service inside a Kubernetes cluster. Think of it as the Kubernetes equivalent of a Homebrew formula, an apt dpkg, or a yum RPM file. To learn how to build one, see the Charts guide.

Repository​

A repository is the place where charts are collected and shared. It works like Perl's CPAN archive or the Fedora Package Database, but for Kubernetes packages. You can find publicly available charts on Artifact Hub, which lists charts from many repositories. For more information about hosting charts, see the Chart Repository guide.

Release​

A release is an instance of a chart running in a Kubernetes cluster. You can install one chart many times into the same cluster, and each installation creates a new release with its own release name. For example, if you want two databases running in your cluster, you can install a MySQL chart twice, and each installation is tracked as a separate release.

To create a release, Helm merges a packaged chart with configuration information. The configuration is a set of values, typically from a values.yaml file.

Architecture​

Helm is a command-line tool that runs on your local machine and talks to the Kubernetes API server.

Helm is built in two distinct parts:

  • The Helm client is the command-line tool for end users. It handles local chart development, manages repositories, manages releases, and sends charts to the Helm library to be installed, upgraded, or uninstalled.
  • The Helm library provides the logic that carries out Helm operations. It creates releases, and it installs, upgrades, and uninstalls charts by interacting with the Kubernetes API server. Because the library is standalone, other clients can reuse the same logic.

The Helm client and library are written in the Go programming language, and the library uses the Kubernetes client library to communicate with Kubernetes over REST and JSON. Helm stores release information in Kubernetes Secrets inside the cluster, so it does not need its own database. Configuration files are written in YAML where possible.

Next Steps​