Deploy and Configure a Nutanix Cluster
Deployment items on NCP-MCI test whether you can take bare-metal nodes to a healthy, monitored cluster: imaging, IP planning, cluster creation, redundancy choices, and the post-build checklist.
Imaging and IP planning
- Foundation images nodes by booting them from a Foundation VM/appliance over IPMI; Foundation Central does the same at scale without a local Foundation VM.
- Plan four addresses per node: IPMI, AHV host, CVM, plus one cluster virtual IP used by Prism Element. The VIP lives on one CVM at a time and follows the CVM if it fails.
Cluster creation
- The create-cluster wizard sets the cluster name, VIP, DNS/NTP, hypervisor (AHV, ESXi or Hyper-V), redundancy factor (RF2/RF3), and container defaults.
- NTP and DNS must resolve correctly before creation - authentication, Prism Central registration and alert email all depend on them.
- RF2 tolerates the loss of one node (FT1); RF3 needs at least five nodes to deliver FT2.
- After creation: apply an LCM firmware baseline, register with Prism Central, confirm Pulse/telemetry, and run NCC.
Worked example
A four-node RF2 cluster is proposed for a workload that must survive the loss of any single node. RF2 on four nodes gives FT1, so the requirement is met - but the surviving three nodes must absorb the failed node's data and CPU, so capacity planning must leave headroom. If the requirement were two simultaneous failures, RF2 would fail and the design would need RF3 on at least five nodes.
Traps
- Confusing the cluster VIP with a CVM IP when configuring firewalls.
- Skipping NTP: time skew breaks the Cassandra/Zeus metadata ring and Prism logins.
- Adding a node with a different AOS build or unsupported hardware model.