DatadogAgentProfile: support to disable APM

[adrienzieba]

Hello,

In my cluster I have node groups that don't need APM. I tried several approaches.

• First, I tried to use 2 datadog agent in the cluster with different toleration to be able to have 2 different configurations. By doing that I stumbled on Can the operator reconcile multiple DatadogAgent objects? #1330 where the operator loops deleting/creating daemonset.
• I switched to use DAP. The setup works for the scheduling of the daemonset across my different node groups. However, I'm not able to disable APM. I tried using environment variables but without success. The problem is located in the trace-agent container that will crashloop when the flag DD_APM_ENABLED is set to false.
• When I compare with a setup where the APM is disabled, I notice that the trace-agent container is completely removed from the daemonset. It is something I cannot do with the current DAP.

Is there any way to disable APM for specific node group?

Example of DAP I tried to use but it makes trace-agent crashlooping

apiVersion: datadoghq.com/v1alpha1
kind: DatadogAgentProfile
metadata:
  name: no-apm
  namespace: datadog
spec:
  config:
    override:
      nodeAgent:
        containers:
          agent:
            env:
              - name: DD_APM_ENABLED
                value: 'false'
          process-agent:
            env:
              - name: DD_APM_ENABLED
                value: 'false'
          trace-agent:
            env:
              - name: DD_APM_ENABLED
                value: 'false'
          system-probe:
            env:
              - name: DD_APM_ENABLED
                value: 'false'
  profileAffinity:
    profileNodeAffinity:
      - key: kube/nodetype
        operator: In
        values:
          - a-node-type

Thanks