Chart Evaluation Checklist

In order for a Helm chart to provide a stable service, it has to satisfy some quality requirements. The service is only as good as the upstream Helm chart.

Checklist Must

All these points are required to ensure optimal compatibility.

  • available via OCI repository

  • compatible with FluxCD

    • test by deploying the chart directly via the helmetica-framework

  • provides image and registry configurations

  • annotations and labels can be passed through to pods and PVCs

  • resource requests and limits are configurable

  • no cluster-scoped resources, or they allow multiple instances (for example by prefixing them), or they can be disabled

  • minor upgrades need no manual steps

Checklist Should

These are nice to have for a better maintainer experience.

  • maps instead of lists in values

  • low values nesting

  • fullnameOverride for predictable resource names

  • options to disable included ingress/gatewayAPI configs

  • values are documented with comments

  • few empty values without a type (null, {}, [])

  • supports existing secrets for credentials

  • provides a metrics endpoint or ServiceMonitor

  • resource requests and limits are configurable

  • runs as non-root with configurable securityContext

  • chart is signed (cosign)

  • upgrade jobs like migrations use Helm hooks

  • follows semver, breaking value changes only in major versions