A workload cluster is in essence a Kubernetes cluster. You can create a cluster of the required specifications to run workloads as needed. However, you cannot store non-temporary data in the node virtual machines of the workload cluster.
When the workload cluster is created, the system automatically sets specific resource reservation values for each node to prevent potential cluster issues, such as nodes going offline, due to excessive resource consumption by cluster addons or workloads. If you need to modify the resource reservation values, you need to contact after-sales engineers for assistance.
| Parameter | Description | Reserved value |
|---|---|---|
| enforce-node-allocatable | Node allocatable constraints | Pods |
| system-reserved cpu | CPU system reservation | 100 m |
| system-reserved memory | Memory system reservation | Related to the memory allocated to the node. See the note below for details. |
| kubeReserved cpu | CPU reservation for system daemons | 200 m |
| kubeReserved memory | Memory reservation for system daemons | 250 Mi |
| eviction-hard | Hard eviction threshold for resources | 300 Mi |
Note:
The reserved value of the system-reserved memory relates to the memory allocated for the node, which is assumed to be N (unit: GiB) during cluster creation:
- When N is 8 or less, the reserved value should be calculated based on N using the following formula: 750Mi + 200Mi × (N - 4). Since the minimum memory for a node is 6 GiB, the reserved value ranges from 1150Mi to 1550Mi.
- When N exceeds 8, the reserved value is fixed at 1550 Mi.
