Check out the system requirements for setting up a vSphere cluster as a Supervisor Cluster with the vSphere networking stack and HAProxy load balancer.
Minimum Compute Requirements
Consider separating the management and workload domain as a best practice. The workload domain will be the Supervisor Cluster where you run workloads. The management domain will contain all the management components such as vCenter Server.
|Minimum Deployment Size
|vCenter Server 7.0
|ESXi hosts 7.0
|Without vSAN: 3 ESXi hosts with 1 static IP per host.
With vSAN: 4 ESXi hosts with at least 2 physical NICs.
The hosts must be joined in a cluster with vSphere DRS and HA enabled. vSphere DRS must be in Fully Automate or Partially Automate mode.
Note: Make sure that the names of the hosts that join the cluster use lower case letters. Otherwise, the enablement of the cluster for Workload Management might fail.
|64 GB per host
|Kubernetes control plane VMs
Minimum Network Requirements
|Static IPs for Kubernetes control plane VMs
|Block of 5
|A block of 5 consecutive static IP addresses to be assigned to the Kubernetes control plane VMs in the Supervisor Cluster.
|Management traffic network
|A Management Network that is routable to the ESXi hosts, vCenter Server, the Supervisor Cluster and load balancer. The network must be able to access an image registry and have Internet connectivity if the image registry is on the external network. The image registry must be resolvable through DNS.
|vSphere Distributed Switch
|All hosts from the cluster must be connected to a vSphere Distributed Switch.
|HAProxy load balancer
|An instance of HAProxy load balancer configured with the vCenter Server instance.
|At least one distributed port group must be created on the vSphere Distributed Switch that you configure as the Primary Workload Network. Depending on the topology of choice, you can use the same distributed port group as the Workload Network of namespaces or create more port groups and configure them as Workload Networks. Workload Networks must meet the following requirements:
Important: The workload network must be on a different subnet than the management network.
|NTP and DNS Server
|A DNS server and NTP server that can be used with vCenter Server.
Note: Configure NTP on all ESXi hosts and vCenter Server .
|Optional. Configure a DHCP server to automatically acquire IP addresses for the management and workload networks as well as floating IPs. The DHCP server must support Client Identifiers and provide compatible DNS servers, DNS search domains, and an NTP server.
The DHCP configuration is used by the Supervisor Cluster. Load balancers may require static IP addresses for Management. DHCP Scopes should not overlap these static IP’s. DHCP is not used for virtual IPs. (VIPs)
|Management Network Subnet
The subnet used for management traffic between ESXi hosts and vCenter Server, and the Kubernetes control plane. The size of the subnet must be the following:
Note: The Management Network and the Workload Network must be on different subnets. Assigning the same subnet to the Management and the Workload networks is not supported and can lead to system errors and problems.
|Management Network VLAN
|The VLAN ID of the Management Network subnet.
|Physical Network MTU
|The MTU size must be 1600 or greater on any network that carries overlay traffic.
|Kubernetes services CIDR range
|/16 Private IP addresses
|A private CIDR range to assign IP addresses to Kubernetes services. You must specify a unique Kubernetes services CIDR range for each Supervisor Cluster.