wip: Helm Charts for the operator

[hessjcg]

This is a work-in-progress draft of the helm chart for the operator.

Here are the configurable values implemented so far:

• Operator Version (default latest. This controls the default proxy version.)
• Number of replicas for the operator deployment
• installCRDs yes/no (Using Method 2 as a separate helm chart)
• Operator pod configuration:
  • resource requirements
  • labels & annotations
  • affinity
• Operator Behavior
  • namespace override
  • namespaces monitored by the webhook
• Support
  • Publish the helm chart
  • Update help documentation
  • Provide examples

A couple of open design questions:

• Is it necessary to have separate helm charts for the CRD and the operator deployment? It seems like we need this because the CRD includes configuration related to the operator's webhook, which depends on the operator namespace. Is there a better way to implement this?
• The build process for the helm chart is a bit convoluted. I'd like to simplify it. Here are the steps as implemented today:
  • run controller-tool to generate the CRD from go
  • run kustomize to generate a single installer yaml with all the resources included.
  • run a custom tool to extract resources into separate helm chart files, and put helm templates in the place of the hardcoded namespace and name.
  • hand-code the operator deployment helm chart.
  • run a custom tool to set the helm chart version and operator versions in Chart.yaml.

Fixes #343

[CoryHagerman]

If you do want to use the CRD in the same chart instead of 2 separate ones you can put it in the crds/ directory instead of templates/

https://helm.sh/docs/chart_best_practices/custom_resource_definitions/