[extended-monitoring] node_exporter doesn't honor kubelet eviction thresholds for nodefs and other system mountpoints in GKE

Preflight Checklist

• I agree to follow the Code of Conduct that this project adheres to.
• I have searched the issue tracker for an issue that matches the one I want to file, without success.

Version

v1.44.4

Expected Behavior

node_exporter should skip nodefs and other kubelet-related filesystems in accordance to Caution clause in Dechouse documentation from being exported.

Actual Behavior

node_exporter in GKE exports corresponding metrics causing Prometheus to alert about space usage in nodefs and other system mountpoints:

Node disk "" on mountpoint "/home/kubernetes/bin" is using more than 70% of storage capacity.
Node disk "" on mountpoint "/home/kubernetes/containerized_mounter" is using more than 70% of storage capacity.
Node disk "" on mountpoint "/home/containerd" is using more than 70% of storage capacity.
Node disk "" on mountpoint "/var/lib/kubelet" is using more than 70% of storage capacity.
Node disk "" on mountpoint "/home/kubernetes/containerized_mounter/rootfs/var/lib/kubelet" is using more than 70% of storage capacity.

Steps To Reproduce

1. Setup Deckhouse in GKE
2. Fill nodefs above 70%
3. Look for alerts in Prometheus

Additional Information

This issue should possibly be fixed in node-exporter's DaemonSet by passing correct mountpoint for GKE.

Logs

No response

Start with modules/340-monitoring-kubernetes/templates/node-exporter/daemonset.yaml around line 81 and the linked extended-monitoring documentation. Reproduce the filesystem alerts in GKE and determine which kubelet-related mountpoints should be excluded. Done means node_exporter no longer exports those system mountpoints for alerts while retaining relevant node filesystem metrics.