Confirm the hardware conditions
Check virtualization support, processors, memory, storage, network interfaces and remote management. Reserve resources for the host OS. Verify address allocation, routing and permission to virtualize against the actual order.
Separate network responsibilities
Plan management, guest, storage and migration traffic explicitly. Bridging, routing and NAT impose different requirements. Create an address allocation table before deployment and restrict management interfaces to trusted sources.
Back up beyond snapshots
Snapshots depend on their underlying storage. A host or storage failure can affect several guests at once. Back up guest data and host configuration and test recovery on a separate system.
Define failure boundaries
Clusters require decisions about quorum, network partitions and storage availability. Two hosts do not automatically provide reliable high availability. Define tolerable downtime before choosing extra nodes, independent backups and manual takeover procedures.