'KubeletCheck' object has no attribute 'kubelet_credentials'

Additional environment details (Operating System, Cloud provider, etc):

• k8s: 1.12.5-gke.5
• datadog-agent: 6.10.1

Steps to reproduce the issue:

1. Run datadog-agent:6.10.1 in a cluster.

Describe the results you received: I’m getting the following warning on startup of any 6.10.1 datadog agents in a Kubernetes cluster on GKE with version 1.12.5-gke.5:

[ AGENT ] 2019-03-08 20:38:29 UTC | WARN | (pkg/collector/py/datadog_agent.go:148 in LogMessage) | (kubelet.py:264) | kubelet check https://10.128.0.4:10250/healthz failed: 'KubeletCheck' object has no attribute 'kubelet_credentials'

Describe the results you expected: Not getting this warning and the failed health check.

Additional information you deem important (e.g. issue happens only occasionally): Looks like the kubelet_credentials value is created here in KubeletCheck.check(..) but used here in KubeletCheck.perform_kubelet_query(..).

The offending method would be KubeletCheck.check(..) call to _perform_kubelet_check here which therefore attempts to call the method which uses those credentials before the method which instantiates them.

[xvello]

Indeed, there is a lifecycle issue in the kubelet check. Although subsequent runs will be OK, the first check run will fail, creating a false-positive on the service check. https://github.com/DataDog/integrations-core/pull/3457 will fix it in 6.11.0

In the meantime, you can set monitors to only trigger on two+ consecutive failed checks, to filter out this blip.

[TheKevJames]

My bad, should have closed this issue – I’ve verified its fixed as of 6.11.0. Thanks again!