Helm

What It Is

Helm is the package manager for Kubernetes. It lets you define, install, and upgrade complex Kubernetes applications using a packaging format called a chart. A chart is a collection of YAML templates that together describe a deployable unit: Deployments, Services, ConfigMaps, Ingress rules, RBAC policies, and anything else your application needs. Values in those templates are parameterised, so the same chart can be installed with different settings for development, staging, and production without duplicating a single line of manifest. Helm tracks every installation as a release, maintains a history of changes, and lets you roll back to any previous revision in a single command.

Why It Matters

Kubernetes YAML is notoriously verbose. A production ready deployment for even a modest service might span several hundred lines across six or seven files, with repeated values, subtle ordering dependencies, and namespace references that must stay consistent across all of them. Managing that by hand across three environments means either maintaining three copies of nearly identical files, which diverge over time, or writing custom templating scripts that nobody except their author can understand.

Helm solves the duplication problem at the right abstraction level. Templates handle the structure; a single values.yaml file handles the variation. A developer adding a new environment variable updates one file and every generated manifest reflects the change. The chart format also enables sharing. The Artifact Hub hosts thousands of community charts for common infrastructure components, from Prometheus and Grafana to Kafka and cert-manager. Adopting a well maintained community chart for an off the shelf component means you inherit years of operational experience encoded in sensible defaults, optional features, and documented configuration options.

For platform teams, Helm charts are the standard delivery vehicle for internal platform capabilities. A shared observability stack, a standard ingress configuration, or a compliant database setup can all be packaged as charts and distributed through a private chart repository, ensuring that every team deploys the approved pattern and gets updates through a single upgrade command.

In Practice

• values.yaml and --set overrides: the default chart configuration lives in values.yaml. Environment-specific overrides go in separate files (values.staging.yaml) or are passed via --set flags. CI/CD pipelines typically use --set image.tag=$GIT_SHA to pin each deploy to a specific build.
• helm upgrade --install: idempotent install or upgrade command used in pipelines. The first run creates the release; subsequent runs update it. Combining this with --atomic means the entire upgrade is rolled back automatically if any pod fails to become ready within a timeout window.
• Release history and rollback: helm history <release> shows every revision deployed, with timestamps and status. helm rollback <release> <revision> restores the cluster to a previous state in seconds. This is meaningfully faster than reverting a commit, pushing, and waiting for a pipeline.
• Subcharts and dependencies: a chart can declare dependencies on other charts in Chart.yaml. Running helm dependency update pulls them in and bundles them alongside the parent. This lets platform teams compose complex multi component applications from reusable building blocks.
• Helmfile: a community tool that treats multiple Helm releases as a single declarative unit, applying them in order with shared values and environment-specific overrides. Useful when a service mesh, an ingress controller, and an application all need to be deployed together with consistent configuration.

Key Insight

Helm is most powerful when a team treats the chart as the product, not the application. Investing time in sensible defaults, clear documentation of every value, and useful named templates compounds over time. Every new environment, every new service instance, every new team that adopts the chart gets the benefit of that investment for free. The chart becomes organisational knowledge made executable.