search

kubereboot / kured

Kubernetes Reboot Daemon
https://kured.dev
Apache License 2.0
2.2k stars 204 forks source link

No prometheus metrics from kured on a control plane node #998

Open amrap030 opened 3 weeks ago

amrap030 commented 3 weeks ago

Hello, I am running kured with Helm in the version 5.5.0 and kured v1.16.0 in my Kubernete Cluster with v1.31.1. I want to use prometheus and grafana to monitor kured, so I enabled the service monitor via the Helm values with the following configuration:

metrics:
    create: true
    namespace: ""
    labels:
      release: kube-prometheus-stack
    interval: 60s
    scrapeTimeout: ""

I have 3 nodes in my cluster, 1 control plane and 2 worker nodes. For some reason, prometheus only catches the metrics of the pods on the worker nodes and not of the pod on the control plane, see image below.

image

If I set hostNetwork: true in the Helm values, it works fine, see image below.

image

However, I don't want to set hostNetwork: true. Do you have an idea what might be the problem here?

evrardjp commented 3 weeks ago

We don't treat the master and workers any differently.

So at first sight, I don't know... Maybe a netpol? @ckotzbauer do you have an idea?