Planning the number of cluster nodes

Planning the number of nodes for the management cluster

Select a suitable AKE deployment mode based on your requirements. In different deployment modes, the number of control plane nodes and worker nodes in the management cluster are different, as shown in the table below.

Deployment mode	Use cases	Node role and count in the management cluster
High availability mode	Suitable for the production environment and can ensure control plane high availability.	3 control plane nodes
		3 worker nodes
Basic mode	Suitable for the development and testing environment and can smoothly upgrade to High Availability Mode.	1 control plane node
		1 worker node

Planning the number of nodes for the workload cluster

Control plane node

There is only one control plane node group in a workload cluster, and the number of nodes in that group is only supported by 1, 3 or 5. You can select the appropriate number of nodes according to the high availability requirements and ensure that the number of physical hosts available in the ACOS cluster that runs the workload cluster is not less than the number of nodes selected.

• 1: The control plane cannot ensure high availability , which means the whole cluster will not be available once the control plane node is exceptional.
• 3: Allows up to 1 control plane node to fail.
• 5: Allows up to 2 control plane nodes to fail.

Worker node

Worker nodes in a VM-based workload cluster

A VM-based workload cluster can have multiple worker node groups, and each worker node in the same group has same resource configurations. The number of nodes can be set to any of the following types:

• Fixed node count: The number of nodes is a fixed number. A fixed number of nodes needs to be set, with a minimum of 1.
• Autoscaling: The number of nodes is automatically adjusted in a range according to workloads and whether pods can be scheduled. The range needs to be specified by setting the smallest number of nodes and the maximum number of nodes, at least 3.

The resource configurations and node types of different worker node groups can be different or the same. You can plan the appropriate number of worker node groups, as well as the node type and quantity within each worker node group, based on your actual needs. Additionally, you can calculate the total number of worker nodes. When calculating the number of nodes, for worker node groups with node type Autoscaling, the number of nodes should be calculated based on the maximum number of nodes.

Worker nodes in a physical-machine-based workload cluster

A physical-machine-based workload cluster can have multiple worker node groups. Within the same worker node group, the model and specifications of physical machines can be different. However, for easier management, it is recommended to group physical machines with similar configurations or certain common characteristics in the same node group. For example, you can group physical machines that require the same model of GPU devices in the same node group.

You can plan the appropriate number of worker node groups and the number of nodes in each worker node group based on your actual business needs, and calculate the total number of worker nodes.