Modernize our Helm chart

[consideRatio]

Our Helm chart could use some love and modernizing - to make it easier to reason about, catch mistakes, and modify (rather than just for the sake of modernizing). This isn't about using a framework or anything like that - just cleaning up.

Directly helpful updates

• Add helm template --validate test so our resources are verified to be valid against a k8s api-server
  • Part of Add schema.yaml for helm3 compliant chart validation #1331.
• Add upgrade tests of helm chart in CI system
  • Add a Helm chart upgrade test to the CI system #1195
  • helm-chart/ci: create an upgrade test #1420
• Add values.schema.json file to our Helm chart so end user config is validated before install/upgrade
  • Add schema.yaml for helm3 compliant chart validation #1331
• Improve test coverage by creating the dind daemonset and the image-cleaner deployment in our CI tests of the entire Helm chart
  • ci: ensure creation of all the various k8s resources #1397
  • helm-chart: improve test coverage and fix image-cleaner bug #1417
• Support modern apiVersion fields when available
  • helm-chart: bump Ingress k8s apiVersion #1407 and helm-chart: fix typo for new Ingress apiVersion format #1416
  • helm-chart: support modern api-version of PDBs #1412
• Make sure our binder pod can terminate quickly (to not wait 30s for no use during a helm upgrade).
  Reference fix for work in docker-image-cleaner: Use tini in Dockerfile to react to SIGTERM properly docker-image-cleaner#6.
  • Verify the situation: it is not reacting to SIGTERM as tried by kubectl delete pod binder... in the staging cluster.
  • image: use tini to react on SIGTERM for quick terminations #1421
• Make our image test build and publish for both amd64 and arm64
  Reference issue for work in z2jh: Publish ARM64 Docker images zero-to-jupyterhub-k8s#2119
  • image: test/build/push for arm64 as well #1422
• Refreeze our image's python dependencies
  • image: refreeze dependencies #1423
• Unpin image's kubernetes client library pinning to a far more modern version
• Don't hardcode resource names, support fullnameOverride and nameOverride config
  See https://zero-to-jupyterhub.readthedocs.io/en/latest/resources/reference.html#fullnameoverride
• Reference z2jh resource names by the z2jh chart's named template helper for this purpose
  See https://zero-to-jupyterhub.readthedocs.io/en/latest/resources/reference.html#fullnameoverride

Misc bugs spotted

• helm-chart: fix inability to omit ingress.hosts, make it optional #1411
• ci: stop running scheduled CodeQL jobs that never fail #1405
• helm-chart: fix missing PDB config maxUnavailable, and default to 1 #1418

Refactoring updates

• This was less work in this than expected, the nested checkboxes are covered by helm-chart: minor refactoring of syntax #1419
  • Use modern with syntax instead of if

    # old syntax - repeating the value needed
    {{- if .Values.somethingX.somethingY }}
    something: {{ .Values.somethingX.somethingY }}
    {{- end }}

    # modern syntax - no need to repeat variable
    # with statements are also conditionals: a truthy value is required
    {{- with .Values.somethingX.somethingY }}
    something: {{ . }}
    {{- end }}

  • Use consistent indentation practices and whitespacing practices.
    Practices I've championed have partially been upstreamed all the way to helm's official documentation.
    • Always whitespace chomp left.
    • Also whitespace chomp right in the first entry of a file or defined template.
    • Use nindent and left whitespace chomp instead of indent to be able to indent lines that have otherwise required 0 indentation to not error for multiline content.
  • Stop trimming new lines after toYaml - it's no longer needed with modern helm that we require already.

      # before
      {{- with .Values.annotations }}
      annotations:
        {{- . | toYaml | trimSuffix "\n" | nindent 4 }}
      {{- end }}

      # after
      {{- with .Values.annotations }}
      annotations:
        {{- . | toYaml | nindent 4 }}
      {{- end }}

• Use port names instead of port numbers in network config

Categorization of PRs

In order to do work on the Helm templates, I wanted to make sure there isn't a great conflict with open PRs. I went through all the open PRs and labelled them based on the type of code that is changed based on the labels code:python, code:html/js/css, code:helm-chart. These are the PRs open about changes to the helm chart.

Only two small changes in two PRs were made besides my own PRs, so I figure it's fine for me to do some refactoring of the templates at this point without causing disruptions to existing PRs.

[consideRatio]

I think most if not all of this is done now. Closing.