StarlingX uses several different types of networks, depending on the size of the system and the features in use.
Available networks include the optional PXE boot network, the internal management network, the cluster host network, the OAM network, and other optional networks for external network connectivity.
If you have a small number of nodes, smaller IP subnets / address ranges should be used to preserve IP addresses that make the network more manageable. When planning the system architecture an (optional) PXE boot, management and cluster-host address is required for each node.
PXE Boot network is not “required” in all scenarios; in some scenarios (e.g. IPv4 and non-vlan-tagged management networks), the management network can be used for PXE booting.
The internal management network is required by all deployment configurations for internal communication.
The cluster host network is required by all deployment configurations to support a Kubernetes cluster. It is used for private container-to-container networking within a cluster. It can be used for external connectivity of container workloads. If the cluster host network is not used for external connectivity of container workloads, then either the OAM port or other configured ports on both the controller and worker nodes can be used for connectivity to external networks.
The OAM network is required for external control and board management access. It can be required for container payload external connectivity, depending on container payload application network requirements.
You can consolidate more than one network on a single physical interface. For more information, see Shared (VLAN or Multi-Netted) Ethernet Interfaces.
Systems with two controllers use IP multicast messaging on the internal management network. To prevent loss of controller synchronization, ensure that the switches and other devices on these networks are configured with appropriate settings.