Promitor on Kubernetes CPU & Memory Consumption

[hmrc87]

Hi,
I am experimenting with Promitor mainly in order to monitor our Servicebus queues and topics. I am scraping a "few" of them by using the dimension EntityName,

I have 6 of these metrics defined like this

 - name: azure_service_bus_active_messages
    description: "Count of active messages in a Queue/Topic."
    resourceType: ServiceBusNamespace 
    azureMetricConfiguration:
      metricName: ActiveMessages
      aggregation:
        type: Average
      dimension:
        name: EntityName
    resources:
      - namespace: infra-northeurope-sbns-dev
        resourceGroupName: infra-northeurope-rg-dev
      - namespace: infra-northeurope-sbns-stage
        resourceGroupName: infra-northeurope-rg-stage
      - namespace: infra-northeurope-sbns-prod
        resourceGroupName: infra-northeurope-rg-prod

-> Scraping interval is set to 5min

The problem is: we see a lot of CPU and Memory usage in the Kubernetes cluster, is this expected? Why doesn't the Memory usage go down after a scraping interval? I was expecting a lot less consumption from the scraper

[tomkerkhove]

The CPU is related to all the scraping (you can see it in network calls as well), however, the memory could be better but I think this relates to #798 which is due to the Azure SDK

[hmrc87]

Thanks for the quick answer, Tom. Can I influence the CPU usage anyhow? Is it the shear amount of metrics, or does it depend on the Azure RessourceType I am scraping? Does using resourceDiscoveryGroups instead of the Dimensions (see my example) help?

[tomkerkhove]

I haven't looked at tweeking it more, I'm afraid there is no way to improve it for now - Sorry!