[BUG] CPUBurst cannot co-work with CPUNormalization

[saintube]

What happened:

Koordlet supports two features CPUBurst and CPUNormalization which can change the cpu.cfs_quota_us/cpu.max of the LS pods and containers. However, these features can not co-work for the cfs_quota adjustment. When both features are enabled, the QoS plugin of the CPUBurst modifies the cfs_quota due to the cfs throttled ratio of pods. In contrast, the runtime hook plugin of the CPUNormalization may overwrite the cfs_quota according to the CPU-normalisation ratio of the node. Then neither of them can achieve their reconciliation.

What you expected to happen:

To make the CPUBurst feature co-work with the CPUNormalization, the CPUBurst QoS plugin should be aware of the CPU-normalization ratio of the node. When both features are enabled, the CPUBurst QoS plugin is responsible for adjusting the cpu.cfs_quota_us/cpu.max according to the scaling ratio and the CPU-normalization ratio, and the runtime hook of the CPUNormalization should stop its reconciliation to avoid conflicts.

How to reproduce it (as minimally and precisely as possible):

Anything else we need to know?:

Environment:

• App version: v1.5.0
• Kubernetes version (use kubectl version):
• Install details (e.g. helm install args):
• Node environment (for koordlet/runtime-proxy issue):
  • Containerd/Docker version:
  • OS version:
  • Kernal version:
  • Cgroup driver: cgroupfs/systemd
• Others:

[saintube]

Hi @saintube I came across this issue and would like to work on resolving it. Could you please assign it to me?

OK.

[nakul-py]

I cannot build or install