Helm chart improvements

[reubenhEROAD]

I am self hosting featurehub, using terraform to install the helm chart on EKS, and Datadog for observability.

I have had success getting everything running but I did have to make a few modifications to the helm chart. I am wondering if anyone else has encountered these items, if there is any way around them, or if they might be addressed in the future so I can use the provided helm chart instead of self hosting:

• To expose the prometheus metrics to the datadog agent, an annotation with a JSON value needs to be added to the pods, shown below. Helm doesn’t like this and the only way to make it work has been to add toJson function to the deployment templates like below. Perhaps a solution could be a seperate podAnnotationsJson value which converts annotation values into JSON using the helm function.

podAnnotations:
    ad.datadoghq.com/featurehub-dacha.checks: |
      {
        "openmetrics": {
          "init_config": {},
          "instances": [
            {
              "openmetrics_endpoint": "http://featurehub-dacha.featurehub:8701/metrics",
              "namespace": "eks.featurehub.dacha",
              "metrics": [".*"]
            }
          ]
        }
      }

{{- range $key, $value := .Values.edge.podAnnotations }}
        {{ $key }}: {{ $value | toJson }}
{{- end }}

• Enabling .Values.*.prometheus.enabled enables the ServiceMonitor resources - this requires installation of the prometheus operator into the cluster and is not needed for my case, so ideally there would be a seperate value to toggle to expose prometheus port on the services and enable the ServiceMonitor
• To correctly tag all of the data collected by Datadog I need to add labels to the deployments and pods. The nats helm chart has values for this statefulSetPodLabels & commonLabels - it would be great if the featurehub helm chart could support this
• The ingress load balancers used by EKS only work with ClusterIP services and not NodePort, so ideally would be able to toggle/choose between the two service types
• And finally I think this is a mistake that no one has picked up yet: the edge & MR deployment templates refer to a value .Values.podAnnotations when it should be .Values.managementRepository.podAnnotations & .Values.edge.podAnnotations

Maybe I could be the first person to encounter these issues. I am happy to try contribute to improve the helm chart (please let me know if there is anything I need to consider before making a PR?)
Thanks 🙂

[reubenhEROAD]

The method used by the NATS helm chart appears to handle JSON annotation values fine, so using this way could be an option:
https://github.com/nats-io/k8s/blob/main/helm/charts/nats/templates/statefulset.yaml#LL42C7-L44C17

[IrinaSouth]

hi @reubenhEROAD ! Thanks for the detailed suggestion for the helm chart improvement, please go ahead and raise a PR in the Helm repo. Cheers!