Helm Architecture: Client and Library
You'll learn how Helm's client-side architecture operates without requiring a server component in your cluster.
How Helm Works Under the Hood
Modern Helm (v3+) uses a client-only architecture that's dramatically simpler than its predecessor. When you run helm install or helm upgrade, you're invoking the Helm client binary on your local machine. This client doesn't need any special server-side component running in your Kubernetes cluster—it talks directly to the Kubernetes API server, just like kubectl does.
Think of Helm as a smart wrapper around your Kubernetes API interactions. The Helm client contains a library that packages all the logic for templating, rendering charts, validating manifests, and tracking releases. When you execute a command, the client:
- Renders templates locally by combining your chart files with values
- Generates Kubernetes manifests (YAML)
- Sends those manifests to the Kubernetes API server via the standard API
- Stores release metadata as Secrets (or ConfigMaps) in your cluster
The Library Component
Helm's architecture separates concerns between the CLI tool (what you interact with) and the Go library that powers it. This library can be embedded in other applications, enabling tools to programmatically manage Helm charts without shelling out to the CLI. The library handles chart loading, dependency resolution, value merging, and template rendering—all the heavy lifting.
No Tiller Anymore
If you've heard of Tiller, forget it. Helm v3 removed this server-side component entirely, eliminating security concerns and simplifying permissions. Now Helm operates with your kubectl credentials, meaning it has exactly the permissions you have—no more, no less.
Key Takeaway: Helm v3 is a client-side tool with a reusable Go library that communicates directly with the Kubernetes API server, storing release state as cluster resources without requiring server-side components.