Host sensor warns continuously when kube-proxy is intentionally absent

What happened?

On a Kubernetes cluster that deliberately runs without kube-proxy (Cilium kube-proxy replacement), enabling the node-agent host sensor emits this warning on every host-sensor interval:

sensor failed kind=KubeProxyInfo error="failed to sense data: failed to locate kube-proxy process: ..."

This is an expected cluster state rather than a sensor failure. The warning creates recurring operational noise and makes warning-based health views less actionable.

Why it happens

NewHostSensorManager unconditionally includes NewKubeProxyInfoSensor whenever host sensing is enabled. KubeProxyInfoSensor.Sense returns an error when it cannot locate the kube-proxy process. The same behavior is present in v0.3.142 and current main.

There is currently one switch for the complete host-sensor manager, so an operator cannot suppress only this inapplicable sensor without also losing OS, kernel, kubelet, open-port, hardening, and CNI data.

Requested behavior

Please support kube-proxy-free clusters without a recurring warning. Either of these contracts would work:

1. Treat an absent kube-proxy process as a supported empty/not-applicable result; or
2. Add a per-sensor exclusion list so operators can disable KubeProxyInfo while retaining the other host sensors.

The first option seems preferable when kube-proxy replacement can be detected reliably. A test should cover a host with no kube-proxy process and prove the periodic sensing loop remains warning-free while the other sensors still run.

Versions checked

• node-agent v0.3.142
• current main as of 2026-09-08

Start with NewHostSensorManager and NewKubeProxyInfoSensor, then follow KubeProxyInfoSensor.Sense into the periodic host-sensing loop. Check how other sensors report empty or not-applicable results and determine which requested contract fits. Done means a host without kube-proxy produces no recurring warning while the other host sensors continue running, with a test covering that case.