Node Resource Reservation Rules

Querying Node Allocatable Resources

Use kubectl to access the cluster and run the following command to view a summary of all nodes. The output lists node names and statuses.

kubectl get nodes

View the total resources and allocatable resources of a specific node. Replace <node_name> with the name of the target node.

kubectl describe node <node_name> | grep Capacity -A 17

Expected output:

Capacity:
  cpu:                2                 # The number of CPU cores on the node
  ephemeral-storage:  102622136Ki       # Total ephemeral storage of the node, in KiB
  hugepages-1Gi:      0
  hugepages-2Mi:      0
  localssd:           0
  localvolume:        0
  memory:             3852580Ki         # Total memory of the node, in KiB
  pods:               20
Allocatable:
  cpu:                1930m             # The number of node's allocatable CPU cores
  ephemeral-storage:  94576560382       # The node's allocatable ephemeral storage, in bytes
  hugepages-1Gi:      0
  hugepages-2Mi:      0
  localssd:           0
  localvolume:        0
  memory:             2632996Ki          # The node's allocatable memory, in KiB
  pods:               20

Relationship Between Node Resource Reservation and Cluster Versions

Node Memory Reservation Rule v1

You can use the following formula to calculate how much memory you should reserve for running containers on a node:

Total reserved amount = Reserved memory for system components + Reserved memory for kubelet to manage pods

For a small-capacity node, adjust the maximum number of pods based on site requirements. Alternatively, when creating a node on the CCE console, you can adjust the maximum number of pods on the node based on the node specifications.

Node Memory Reservation Rule v2

The total reserved memory of a CCE node using memory reservation rule v2 includes memory allocated for system components and Kubernetes.

Reserved memory includes basic and floating parts. For system components, the floating memory depends on the node specifications. For Kubernetes, the floating memory depends on the number of pods on a node.

By default, CCE automatically calculates the reserved memory based on the rules listed in Table 3. You can specify the reserved memory by configuring parameters system-reserved-mem and kube-reserved-mem in the node pool settings. For details, see Configuring a Node Pool.

Node CPU Reservation Rules

Node Data Disk Reservation Rules

CCE uses Logical Volume Manager (LVM) for disk management. LVM creates a metadata area on a disk to store logical and physical volume information, occupying 4 MiB of space. Therefore, the actual available disk space of a node is the disk size minus 4 MiB.

Helpful Links

For details about how to configure resource reservation and eviction thresholds, see Configuring a Node Pool.