Memory consumption on control-plane nodes

[pat-s]

I am wondering a bit about the memory consumption of the control plane nodes and their reported values from different tools.

Scenario

• 3 control planes spread across fsn1, nbg1, hel1 running on cax11
• Klipper LB

Now, kubectl top no returns

auto-large-amd64-15019bef2d9487e9   622m         8%     9166Mi          61%
auto-large-arm64-26c5fcbb872dbebf   339m         4%     8245Mi          55%
control-plane-fsn1-bns              142m         9%     2161Mi          107%
control-plane-hel1-czh              112m         7%     1924Mi          95%
control-plane-nbg1-fax              143m         9%     2635Mi          130%

This made me worry a little, especially the value for control-plane-nbg1-fax.

I ssh'ed into the node and saw that htop reported a total memory consumption of 1.99/3.7 GB (which is totally fine).
Further, I saw that k3s-agent claimed 1.7 GB of Memory (46%).

Questions

• Is a consumption of 1.7 GB for k3s-agent expected and fine?
• Why does kubectl top no return 130% memory allocation on the node? I.e. shouldn't be a lot more memory available?
• Should I care/be concerned about the memory consumption on the control nodes at all?