Cloudwatch agent scrap prometheus metrics

I run pods on a kubernetes running inside EKS. I have Prometheus installed on the cluster. I wish to export to Cloudwatch metrics some metric comming from Prometheus. So I follow this guide.

When comes time to modify my cwagent config, I have this in the current one:

 {
  "agent": {
    "region": "${log_region}"
  },
  "logs": {
    "metrics_collected": {
      "kubernetes": {
        "cluster_name": "${cluster_name}",
        "metrics_collection_interval": 60
      }
    },
    "force_flush_interval": 5
  }
}

Since I still want the base kubernetes metric scrapping working, I was planning on letting it and just adding the "kubernetes" section. Giving me this:

 {
  "agent": {
    "region": "${log_region}"
  },
  "logs": {
    "metrics_collected": {
      "kubernetes": {
        "cluster_name": "${cluster_name}",
        "metrics_collection_interval": 60
      },
      "prometheus": {
        "prometheus_config_path": "/etc/prometheusconfig/prometheus.yaml",
        "emf_processor": {
          "metric_declaration_dedup": false,
          "metric_declaration": [
            ...
          ]
        }
      }
    },
    "force_flush_interval": 5
  }
}

But if I do so, when starting the agent I get an error stating that I can not have "Prometheus" and "Kubernetes" at the same time:

error : "feature kubernetes, ecs, prometheus are mutually exclusive"

So I am not sure how I should proceed, if should I make my configuration different to allow multiple scrapper?
Should I totally replace the export of metrics from Kuebernetes to be replaced by the one from Prometheus ? If so, is there a way to easily do it, or at least find the list of what metrics were here it the first place?

Solution

So the solution is actually to create a second agent. I have one agent working as daemon set on all my nodes, scrapping Kub metrics. And I create a deployement specifically for an CWAgent working with Kubernetes.