Opscraft
Helm & Kustomize · Lesson 6 of 7 · 3 min read

Choosing and combining

Decision guide

  • Third-party software → its official Helm chart, pinned version, values in Git, applied by Flux HelmRelease / Argo CD.
  • Your services, few environments, simple differences → Kustomize base + overlays.
  • Many similar services → one internal Helm chart with good defaults (or a Kustomize component library); each service has a values file/overlay.
  • Need to tweak a chart without forking → Flux post-renderers or Kustomize helmCharts + patches.
  • Need hooks/rollbacks outside GitOps → Helm.
Enlarge diagramThird-party softwareits official Helm chart, pinned, values in GitYour services, few environmentsKustomize base + overlaysMany similar servicesone internal Helm chart + values per serviceTweak a chart without forkingpost-renderer or Kustomize patches
Figure 1. Choosing between Helm and Kustomize.
Swipe sideways to see all of it, or tap to open it full screen.

Pitfalls to mention in interviews

  • Template sprawl — charts with hundreds of ifs are hard to review; keep charts opinionated and use values.schema.json.
  • Always helm template/kustomize build and diff in PRs — reviewers should see rendered YAML changes.
  • Pin chart versions; latest or floating ranges in prod surprise you.
  • Immutable fields (selectors, Service clusterIP, StatefulSet volumeClaimTemplates) — changes need recreate/migration plans.
  • commonLabels changing selectors on existing Deployments breaks applies — use labels with includeSelectors: false.
  • Helm CRDs aren't upgraded automatically.
  • Drift: manual kubectl edits are overwritten by the next Helm upgrade or GitOps reconcile — by design.