The CRD for the operator uses apiVersion: apiextensions.k8s.io/v1beta1 and this is officially deprecated in k8s 1.22
https://kubernetes.io/docs/reference/using-api/deprecation-guide/#customresourcedefinition-v122.

This will cause this CRD are related operator to not deploy into any kubernetes cluster v1.22 or higher.

Need to update the apiextensions to v1 and update related schema, validations.

Prometurbo will created proxy entities for those deleted pods as Prometheus server will still have the metrics available. Prometurbo currently has no way to figure out if pods are deleted.

At server side, the proxy entities of the deleted pods will not be stitched with kubeturbo. It's not desired to show them in UI. So, need to investigate how to remove the proxy entities in SDK framework if they are not stitched at server side.

Currently, when added one addon to support a new exporter, we need to write code for the exporter.

After looking into some examples as Cassandra and XL server, it seems that we can have one generic addon that can allow users to specify the queries to get the metrics and the mapping from the metrics to the corresponding commodities. For example, user can simply add following to a configuration file:

promethuest-query-1 RESPONSE_TIME
promethuest-query-2 TRANSACTION

and prometurbo will take the settings and query the promethues server for the metrics and generate response time and transaction commodities as in our current prometurbo mechanism.

The above is a simplified version but as discussed with the team, this generic addon should be doable and shouldn't be too complicated. Need further investigation and think an optimal way for the framework.

The vApp (service) commodities of transaction and response time doesn't get updated if the service has more than one ContainerPod instances associated.

Some suspect is on stitching at Server side. Need further investigation about the root cause.

The operator is looking for global values for prometheusServers
https://github.com/turbonomic/prometurbo/blob/master/deploy/prometurbo/templates/configmap-prometurbo.yaml

  name: prometurbo-config-{{ .Release.Name }}
data:
  prometheus.config: |-
    # Prometheus servers and metrics to scrape
    servers:
{{- if .Values.prometheusServers }}
{{ tpl .Values.prometheusServers . | indent 6 }}
{{- end }}

but there is no value defined for it in the config
deploy/prometurbo-operator/config/crd/bases/charts.helm.k8s.io_prometurboes.yaml

The outcome of this is a default configmap with
https://github.com/turbonomic/prometurbo/blob/master/deploy/prometurbo/values.yaml

prometheusServers: |
  server1:
    # The URL of the prometheus server
    url: http://Prometheus_Server_URL
    # The list of supported exporters for the above prometheus server
    exporters:
      - cassandra
      - istio
      - jmx-tomcat
      - node
      - redis
      - webdriver

Currently, capacities for Transaction and ResponseTime are hard-coded.

We need to support configure capacity when deploying Prometurbo as it usually depends on applications. Moreover, we need to allow different capacities for difference services.