VxRail supports a 2-node vSAN cluster configuration. With its small footprint, this configuration can be an appropriate, cost-effective solution for locations with limited space and workload requirements. The configuration must be a new deployment, which means existing clusters cannot use node removal to convert to a 2-node cluster configuration. However, a 2-node cluster configuration can be expanded through the addition of new nodes, up to the 64-node cluster limit. Users can still benefit from VxRail automated life cycle management.
The 2-node vSAN cluster on VxRail can be deployed in a switch configuration or a direct-connect configuration. In a switch configuration, all ports are connected to the switch. In a direct-connect configuration, the ports for vSAN and vMotion traffic are directly connected. A witness provides quorum for the cluster. The witness is a virtual appliance installed on an ESXi host that must reside outside the 2-node cluster—that is, in another data center or a physical host in the same rack or location. The witness has individual connections to both nodes, which requires VLANs to separate witness management traffic from vSAN traffic. The configuration only supports mirroring (FTT=1). The witness host is used as the tiebreaker. Each node and the witness are individual fault domains, for a total of three in the cluster.
A special workflow in the first-run experience is used to deploy the 2-node cluster. The workflow includes the setup of the witness appliance and witness traffic separation. The configuration must use a customer-supplied vCenter for management. Because the cluster has only two data nodes, users must be cognizant of the cluster load to prevent data unavailability in case a failure causes a single node to have to service the entire cluster workload.