Scylla-operator cluster nodes utilizes 100% of CPU even with no load

[Psaawkel]

Hello, 
I am running Scylla cluster in kubernetes, managed by Scylla-operator and noticed it utilizes CPU to the pod defined limits even if not loaded with requests when standalone pod with similar settings does not.

I also noticed when I installing cluster (kubernetes is creating first pod, and then, after 3 minutes the second) the first pod is starting and utilizing normally at 2-3% of CPU, and when the second node is created and started - both uses 100% of CPU limit assigned to pod.



I already set up an issue here with details (I collected perf traces as it could be useful for project development):
https://github.com/scylladb/scylladb/issues/12774

Anyone had this issue?

Thanks in advance.