We deploy one application to dozens of customer-owned Kubernetes clusters, using one chart and a values file per customer. The templates are filling up with conditionals for ingress, storage, registries and Kubernetes versions. How do you keep this maintainable without creating a separate chart for everyone?
How do you keep one Helm chart manageable across lots of customer clusters?
Answers
Start by separating configuration differences from structural differences. Keep common defaults in a base values file, then layer customer overrides on top. An optional component may need a conditional, but a registry address or storage class usually belongs in values rather than a branching template.
Helmfile or Kustomize can help assemble those layers. Keep the number of supported configurations bounded, though. Moving every arbitrary customer request into an overlay still leaves you maintaining every arbitrary customer request.
Reply to this answer
Consider drawing a harder line around networking and storage. Expose a ClusterIP service and let the customer manage ingress, certificates and routing. For persistent storage, accept an existing claim rather than trying to cover every storage setup in the chart.
That makes the integration contract clearer, but it also means customer teams have more resources to provision and maintain. Document what the application expects from those resources.
Reply to this answer
Treat the chart like a versioned product, not just a pile of deployment YAML. Define the combinations you support, use semantic versioning, and run automated checks against representative values and Kubernetes versions. Communicate breaking changes and give customers an upgrade path.
For API differences, Helm exposes .Capabilities.KubeVersion and .Capabilities.APIVersions.Has, so some customer-specific version flags can disappear. Keep in mind that helm template does not query a cluster. In CI, pass the relevant --kube-version and --api-versions values when testing those branches.
Reply to this answer
Another option is to maintain an upstream chart with sensible defaults and the features you actually support, then let customers own their customizations in their own repositories. They can pull updates and resolve any conflicts that affect their changes.
That reduces central maintenance, but it pushes upgrade work onto customers. It only works if that ownership is clear and their teams are willing to take it on.
Reply to this answer
If the templates themselves are the painful part, consider whether Kustomize overlays would be easier to maintain than Helm templating for your setup. A central configuration can still point to common resources while customer-specific patches live separately.
It is a different tradeoff, not an automatic cure. You still need to manage variants and check that patches keep working as the base changes.
Reply to this answer
Add your answer
Answers are reviewed before publication.
2p~vexluna!:
Known configuration profiles with explicit overrides can also make bad or undersized customer inputs easier to spot before deployment.