Prometheus set up in CRD refers to the operator deployed in the same namespace

[rgordill]

Is your feature request related to a problem? Please describe.
When you install prometheus operator using OLM, it is installed in a common namespace (operators, or openshift-operators). However, related with the code in https://github.com/argoproj-labs/argocd-operator/blob/master/pkg/controller/argocd/prometheus.go#L177, it refers to prometheus-k8s serviceaccount, which should be in the same namespace and not in the global one.

Describe the solution you'd like
A way to deploy prometheus which is not coupled to where the prometheus operator is deployed.

Describe alternatives you've considered
Copying the prometheus-k8s SA in the namespace, and adding it to the clusterrolebinding.

Additional context
None.

[iam-veeramalla]

Hi @rgordill , Can you please provide more information on what is the issue that you are running into and steps to reproduce ? Thanks in Advance :)

[rgordill]

Hi @iam-veeramalla.

Yes, to reproduce the issue do the following:

• Deploy a prometheus operator in a namespace with cluster-wide scope
• Deploy argocd operator in any namespace (ex. argocd).
• Deploy an argocd-server with prometheus enabled=true

It will not spin up the prometheus server to gather the metrics. You need to deploy the prometheus operator in the same namespace to get it working.

You can easily deploy a prometheus operator cluster-wide if you deploy OLM (Operator Lifecycle Manager) and prometheus from there. (See https://operatorhub.io/operator/prometheus, and follow the install instructions, clicking in the "install" button).

Thanks.