tencent cloud

TKE clusters occupy node resources to run add-ons (such as kubelet, kube-proxy, and runtime). Therefore, the total number of node resources and the number of allocable resources in a cluster may differ from each other. This document describes the policies and notes in terms of node resource reservation in TKE clusters so that you can set reasonable numbers of requested resources and limited resources for Pods when deploying an application.

Policy for Calculating Allocable Node Resources

Calculation formula

Allocable = Capacity - Reserved - Eviction - Threshold

Node CPU reservation rules

Node memory reservation rules

Note：

You can use custom kubelet parameters to modify kube-reserved for node resource reservation. We recommend you reserve sufficient CPU and memory resources for add-ons to ensure node stability.

Viewing Allocable Node Resources

Run the following command (replace NODE_NAME with the actual node name) to check the allocable node resources in a cluster. The output result contains Capacity and Allocatable fields, along with measurements of CPU, memory, and temporary storage.

kubectl describe node NODE_NAME | grep Allocatable -B 7 -A 6

Notes

• The reservation policy automatically takes effect for K8s v1.16 or later and nodes created after June 24, 2022, without no manual configuration required.
• To ensure your business stability, the reservation policy won't take effect for existing nodes. This is because allocable resources may become fewer based on the calculation method, which means possible node eviction for nodes requiring a large number of resources.
• If you want to apply the reservation policy to existing nodes, remove them from the cluster without termination and then add them in the TKE console. In this case, they become newly added nodes subject to the policy by default.