In the CCE cluster, some necessary system components and Kubernetes components should run in the node initialization to enable the node to be successfully incorporated into the cluster and provide services normally. Therefore, some system resources should be reserved for these components. You can see that the total node resources and the distributable resources of the node in Kubernetes have differences. The higher the specification of the node is, the more the containers which may be deployed in the node are, and the more the resources which should be reserved are.

In order to ensure the stability of nodes, some resources can be reserved in the CCE cluster node to the relevant components of Kubernetes (kubelet, kube-proxy, docker, etc.). The computational formula for the finally distributable resources to the node is as below:

Allocatable = Capacity - Reserved - Eviction Threshold

Total distributable resources of the node = total resources - reserved value - eviction threshold.

Eviction configuration

--eviction-hard=memory.available<5%,nodefs.available<10%,imagefs.available<10%% \ 
--eviction-soft=memory.available<10%,nodefs.available<15%,imagefs.available<15%% \ 
--eviction-soft-grace-period=memory.available=2m,nodefs.available=2m,imagefs.available=2m \ 
--eviction-max-pod-grace-period=30 \ 
--eviction-minimum-reclaim=memory.available=0Mi,nodefs.available=500Mi,imagefs.available=500Mi 

Reserved value of memory

• Within 1G : 255Mi
• 1-4G: 1Gi
• 4-8G: 1Gi + (X-4)* 0.2Gi
• 8-16G: 1.8Gi + (X-8)* 0.1Gi
• 16-128G: 2.6Gi + (X-16)* 0.06Gi
• 128G above : 3.56Gi + （X - 128）* 0.02Gi

Reserved value of CPU

• 1 core: 60m
• 2 cores: 70m
• 3 cores: 75m
• 4 cores: 80m
• Greater than 4 cores: 80m + (X-4)* 2.5m

Plug-in assignment

When creating a cluster, CCE installs some plug-ins in the kube-system name space. Certain resources are distributed to these plug-ins, and the resources occupy the whole cluster dimension, rather than the node dimension, including: