Counting cap_perform capability in capsh command output and getting 0

root@honda:~# capsh --print | grep cap_perform | wc -l

You’ve got a typo in there, you want cap_perfmon

root@honda:~# uname -a
Linux honda 5.10.0-0.deb10.16-arm64 #1 SMP Debian 5.10.127-2~bpo10+1 (2022-07-28) aarch64 GNU/Linux
root@honda:~# uname -r
root@honda:~# ```

5.10 kernel should support cap_perfmon from what I read, but this might not be true for arm64 ?
And if so, how can I make kube-proxy work without cap_perfmon? Does it really need it…?

I haven’t seen this, but sounds like it, and has a possible fix

No luck. I will try just upgrade cluster step by step to latest version.

But this one feels tricky… :disappointed: