The container’s CpusetCpus information needs to be added to the CRI Container definition

[chenyw1990]

What happened:

If the K8s runs on bare metal, hundreds of containers run on a node, and the reconcileState method of CPUManager is executed every 10 seconds.
reccileState will call the updateContainerCPUSet method for each container to update the cpusetcpus of the container,
This cause too many cri requests to container runtime, and the cpu usage of container runtime will be high because of too many cri request.

What you expected to happen:

add the cpusetcpus to the Container definition of the CRI. reconcileState call updateContainerCPUSet only when the cpusetcpus of container changes.

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

Anything else we need to know?:

Environment:

• Kubernetes version (use kubectl version): v1.19.4
• Cloud provider or hardware configuration:
• OS (e.g: cat /etc/os-release):
• Kernel (e.g. uname -a):
• Install tools:
• Network plugin and version (if this is a network-related bug):
• Others:

[chenyw1990]

/sig node

[chenyw1990]

@dims

[zouyee]

/assign

[JornShen]

@SergeyKanzhelev @Random-Liu

[SergeyKanzhelev]

Looks like a legit scalability concern.

/triage accepted
KEP: kubernetes/enhancements#693

Perhaps, needs to be added as a blocker for topology manager GA

[SergeyKanzhelev]

CC: @klueska wdyt?

[klueska]

This is a known issue, we just haven't found the time to work on it. I don't think we need to add anything to the Container definition of the CRI. All of the information needed to skip the update is already available.

[SergeyKanzhelev]

This is a known issue, we just haven't found the time to work on it. I don't think we need to add anything to the Container definition of the CRI. All of the information needed to skip the update is already available.

created kubernetes/enhancements#2623 so it is not lost on graduation