G7: Lint and validate the Helm charts #55

Open
opened 2026-11-09 17:06:03 +00:00 by claude · 1 comment
Member

Add helm lint and rendered-manifest validation with kubeconform to helm-jamulus, helm-mailman3, and helm-pixelfed. Set the Kubernetes version to the cluster's.

Acceptance: each repo has a passing workflow, and a broken template fails it.

Depends on: G2.

Add `helm lint` and rendered-manifest validation with kubeconform to helm-jamulus, helm-mailman3, and helm-pixelfed. Set the Kubernetes version to the cluster's. **Acceptance**: each repo has a passing workflow, and a broken template fails it. Depends on: G2.
claude added this to the CI/CD 2: Quality gates milestone 2026-11-09 17:06:03 +00:00
claude added the cicd label 2026-11-09 17:06:03 +00:00
claude added this to the CI/CD Roadmap project 2026-11-09 17:07:33 +00:00
Author
Member

Prepared ahead of S8 and G1: committed and tested locally, not yet run on the runner.

What's committed

  • Reusable workflow (84dbed9): reusable-helm-lint.yaml now lints and renders each values set, not only the defaults. New input render-defaults (default true) for charts whose templates required values the defaults leave empty. The job now fails if a values set renders no resources. While testing I found kubeconform passes on empty input and silently skips files without a .yaml extension, so the render goes to a named rendered.yaml. docs/ci-cd/adding-ci.md documents the ci/ placeholder-values convention and this rollout.
  • DevOps/helm-jamulus@410fd6a: caller .gitea/workflows/lint.yaml, defaults.
  • DevOps/helm-mailman3:
    • @ada1ff3 fixes a real bug. 5facaf7 ("Fixed DATABASE_URL", April) used {{ include 'mailman3/fullname' }}, which is a template parse error, so the chart on main couldn't be rendered or installed with any values. It's now {{ .Release.Name }}-postgresql, the Bitnami subchart's Service name. For the deployed release name mailman3, that's mailman3-postgresql, as intended. I checked the rendered output for release names mailman3 and ci-render. mailman3.fullname would only have matched when the release name contains "mailman3". v1.0.8, which the role deploys, is unaffected, and I didn't cut a release.
    • @73930f7: mailman3/ci/minimal-values.yaml (SQLite, postfix sidecar) and mailman3/ci/postgresql-values.yaml (the deployed shape: bundled PostgreSQL, external SMTP, ingress with TLS). Both are placeholders only. The caller uses render-defaults: false, because the defaults leave admin and the keys empty.
  • DevOps/helm-pixelfed: caller on branch ci/helm-lint (8d156db), not main. It uses upstream's test-values/postgresql-plain.yaml. Once it's on main, Gitea uses .gitea/workflows/ instead of upstream's .github/workflows/, so the hourly Renovate job that fails every run stops too. That's the open decision in #46, so merging is Eric's call. https://gitea.eom.dev/DevOps/helm-pixelfed/pulls/new/ci/helm-lint

All callers are manual-only (workflow_dispatch), with the push/PR trigger in the header, plus concurrency and contents: read. The pushes started no runs; I checked all three repos.

Tested on the workstation. I extracted the workflow's run: step and ran it in alpine/helm:3.19.0 with kubeconform 0.8.0, for Kubernetes 1.31.2:

Case Result
jamulus, defaults pass, 2 resources valid
mailman3, both ci/ values files pass, 7 and 15 resources valid (subchart pulled from Bitnami's OCI registry)
pixelfed, upstream test values (subcharts vendored) pass, 27 resources valid
mailman3, defaults only fails in helm lint (missing admin)
mailman3 templates at 5facaf7 fails in helm lint (parse error)
chart with no templates fails: "rendered no resources"
Service with a string port fails in kubeconform
render-defaults: false and no values files fails with a clear message

actionlint 1.7.12 and yamllint pass on all four workflow files and the values files.

Not tested: a run on the runner, the cross-repo uses: (the G2 acceptance check), or the ci-tools image's helm 3.22.1. Locally I used helm 3.19.0.

Remaining before close

  1. S8 (#41) passes and G1 (#49) publishes the image.
  2. Switch the triggers, run each once, and show a broken template failing on the runner (in a cicd-test-* copy).
  3. pixelfed: Eric decides on #46, then merge ci/helm-lint.

— eom-cicd-roadmap

**Prepared ahead of S8 and G1: committed and tested locally, not yet run on the runner.** **What's committed** - **Reusable workflow** (84dbed9): `reusable-helm-lint.yaml` now lints *and* renders each values set, not only the defaults. New input `render-defaults` (default `true`) for charts whose templates `required` values the defaults leave empty. The job now fails if a values set renders no resources. While testing I found kubeconform passes on empty input and silently skips files without a `.yaml` extension, so the render goes to a named `rendered.yaml`. `docs/ci-cd/adding-ci.md` documents the `ci/` placeholder-values convention and this rollout. - **DevOps/helm-jamulus@410fd6a**: caller `.gitea/workflows/lint.yaml`, defaults. - **DevOps/helm-mailman3**: - **@ada1ff3 fixes a real bug.** 5facaf7 ("Fixed DATABASE_URL", April) used `{{ include 'mailman3/fullname' }}`, which is a template parse error, so the chart on `main` couldn't be rendered or installed with any values. It's now `{{ .Release.Name }}-postgresql`, the Bitnami subchart's Service name. For the deployed release name `mailman3`, that's `mailman3-postgresql`, as intended. I checked the rendered output for release names `mailman3` and `ci-render`. `mailman3.fullname` would only have matched when the release name contains "mailman3". **v1.0.8, which the role deploys, is unaffected, and I didn't cut a release.** - @73930f7: `mailman3/ci/minimal-values.yaml` (SQLite, postfix sidecar) and `mailman3/ci/postgresql-values.yaml` (the deployed shape: bundled PostgreSQL, external SMTP, ingress with TLS). Both are placeholders only. The caller uses `render-defaults: false`, because the defaults leave `admin` and the keys empty. - **DevOps/helm-pixelfed**: caller on branch **`ci/helm-lint`** (8d156db), **not `main`**. It uses upstream's `test-values/postgresql-plain.yaml`. Once it's on `main`, Gitea uses `.gitea/workflows/` instead of upstream's `.github/workflows/`, so the hourly Renovate job that fails every run stops too. That's the open decision in #46, so merging is Eric's call. https://gitea.eom.dev/DevOps/helm-pixelfed/pulls/new/ci/helm-lint All callers are **manual-only** (`workflow_dispatch`), with the push/PR trigger in the header, plus `concurrency` and `contents: read`. The pushes started no runs; I checked all three repos. **Tested on the workstation.** I extracted the workflow's `run:` step and ran it in `alpine/helm:3.19.0` with kubeconform 0.8.0, for Kubernetes 1.31.2: | Case | Result | |---|---| | jamulus, defaults | pass, 2 resources valid | | mailman3, both `ci/` values files | pass, 7 and 15 resources valid (subchart pulled from Bitnami's OCI registry) | | pixelfed, upstream test values (subcharts vendored) | pass, 27 resources valid | | mailman3, defaults only | fails in `helm lint` (missing `admin`) | | mailman3 templates at 5facaf7 | fails in `helm lint` (parse error) | | chart with no templates | fails: "rendered no resources" | | Service with a string port | fails in kubeconform | | `render-defaults: false` and no values files | fails with a clear message | `actionlint` 1.7.12 and `yamllint` pass on all four workflow files and the values files. **Not tested:** a run on the runner, the cross-repo `uses:` (the G2 acceptance check), or the ci-tools image's helm 3.22.1. Locally I used helm 3.19.0. **Remaining before close** 1. S8 (#41) passes and G1 (#49) publishes the image. 2. Switch the triggers, run each once, and show a broken template failing on the runner (in a `cicd-test-*` copy). 3. pixelfed: Eric decides on #46, then merge `ci/helm-lint`. — eom-cicd-roadmap
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: DevOps/software-infrastructure#55