#Creating an On-Premise Cluster

All optional versions are rigorously tested for stability and compatibility.

Recommendation: Choose the latest version for optimal features and support.

Containerd is provided as the default container runtime.

Supports three modes: IPv4 single stack, IPv6 single stack, IPv4/IPv6 dual stack.

Note: If you select dual stack mode, ensure all nodes have correctly configured IPv6 addresses; the network protocol cannot be changed after setting.

IP Address / Domain: Enter the pre-prepared load balancer VIP. Although the field label also offers a domain, the cluster endpoint does not support domain access — always enter the IP address.

Self-built VIP: Disabled by default. Only enable if you haven't provided a LoadBalancer. When enabled, the installer will automatically deploy keepalived for software load balancing support.

External Access: Enter the externally accessible address prepared for the cluster when it's not in the same network environment as the global cluster.

The name of the host network interface device used by the cluster network plugin.

Note:

• When selecting Underlay transmission mode for the Kube-OVN default subnet, you must specify the network interface name, which will be the default NIC for bridge networking.
  - The platform's network interface traffic monitoring by default recognizes traffic on interfaces named like eth.|en.|wl.|ww.. If you use interfaces with different naming conventions, please refer to Collect Network Data from Custom-Named Network Interfaces after cluster onboarding to modify the relevant resources and ensure the platform can properly monitor network interface traffic.

You can choose to use either the node IP or hostname as the node name on the platform.

Note: When choosing to use hostname as the node name, ensure that the hostnames of nodes added to the cluster are unique.

Add nodes to the cluster, or Recovery from draft temporarily saved node information. See the detailed parameter descriptions for adding nodes below.

Supports Prometheus and VictoriaMetrics.
When selecting VictoriaMetrics as the monitoring component, you must configure the Deploy Type:
- Deploy VictoriaMetrics: Deploys all related components, including VMStorage, VMAlert, VMAgent, etc.

- Deploy VictoriaMetrics Agent: Only deploys the log collection component, VMAgent. When using this deployment method, you need to associate with a VictoriaMetrics instance already deployed on another cluster in the platform to provide monitoring services for the cluster.

Select nodes for deploying cluster monitoring components. Supports selecting compute nodes and control plane nodes that allow application deployment.

To avoid affecting cluster performance, it's recommended to prioritize compute nodes. After the cluster is successfully created, monitoring components with storage type Local Volume will be deployed on the selected nodes.

Control Plane Node: Responsible for running components such as kube-apiserver, kube-scheduler, kube-controller-manager, etcd, container network, and some platform management components in the cluster. When Application Deployable is enabled, control plane nodes can also be used as compute nodes.

Worker Node: Responsible for hosting business pods running on the cluster.

The IPv4 address of the node. For clusters created in internal network mode, enter the node's private IP.

Valid when the cluster has IPv4/IPv6 dual stack enabled. The IPv6 address of the node.

Valid when Node Type is Control Plane Node. Whether to allow business applications to be deployed on this control plane node, scheduling business-related pods to this node.

The IP address that can connect to the node when accessing it via SSH service.

If you can log in to the node using ssh <username>@<node's IPv4 address>, this parameter is not required; otherwise, enter the node's public IP or NAT external IP to ensure the global cluster and proxy can connect to the node via this IP.

Enter the name of the network interface used by the node. The priority of network interface configuration effectiveness is as follows (from left to right, in descending order):

Kube-OVN Underlay: Node NIC > Cluster NIC

Kube-OVN Overlay: Node NIC > Cluster NIC > NIC corresponding to the node's default route

Calico: Cluster NIC > NIC corresponding to the node's default route

Flannel: Cluster NIC > NIC corresponding to the node's default route

Note: When creating a cluster, bridge network configuration is not supported; this option is only available when adding nodes to a cluster that already has Underlay subnets created.

Select an existing Add Bridge Network. If you don't want to use the bridge network's default NIC, you can configure the node NIC separately.

SSH service port number, e.g., 22.

SSH username, needs to be a user with root privileges, e.g., root.

Whether to access the node's SSH port through a proxy. When the global cluster cannot directly access the node to be added via SSH (e.g., the global cluster and workload cluster are not in the same subnet; the node IP is an internal IP that the global cluster cannot directly access), this switch needs to be turned on and proxy-related parameters configured. After configuring the proxy, node access and deployment can be achieved through the proxy.

Note: Currently, only SOCKS5 proxy is supported.

Access URL: Proxy server address, e.g., 192.168.1.1:1080 .
Username: Username for accessing the proxy server.

Password: Password for accessing the proxy server.

Authentication method and corresponding authentication information for logging into the added node. Options include:

Password: Requires a username with root privileges and the corresponding SSH password.
Key: Requires a private key with root privileges and the private key password.

Saves the currently configured data in the dialog as a draft and closes the Add Node dialog.

Without leaving the Create Cluster page, you can select Restore from draft to open the Add Node dialog and restore the configuration data saved as a draft.

Note: The data restored from the draft is the most recently saved draft data.

kubeletExtraArgs, additional configuration parameters for Kubelet.

Note: When the Container Network's Node IP Count parameter is entered, a default Kubelet Parameter configuration with the key max-pods and a value of Node IP Count is automatically generated. This sets the maximum number of pods that can run on any node in the cluster. This configuration is not displayed in the interface.

Adding a new max-pods: maximum number of runnable pods key-value pair in the Kubelet Parameters area will override the default value. Any positive integer is allowed, but it's recommended to use the default value (Node IP Count) or enter a value not exceeding 256.

controllerManagerExtraArgs, additional configuration parameters for the Controller Manager.

schedulerExtraArgs, additional configuration parameters for the Scheduler.

apiServerExtraArgs, additional configuration parameters for the APIServer.

publicAlternativeNames, APIServer access addresses issued in the certificate. Only IPs or domain names can be entered, with a maximum of 253 characters.

Cluster annotation information, marking cluster characteristics in metadata in the form of key-value pairs for platform components or business components to obtain relevant information.