Section 3 Describe Cluster Configuration and Maintenance

Name servers, NTP, SMTP and SNMP are configured in Prism Element (gear icon → Settings), and are also visible from Prism Central. NTP matters most: hosts and CVMs must agree on time, and the NCC check check_ntp flags drift or unreachable servers.

ncc health_checks network_checks check_ntp

Fault tolerance (FT) is how many simultaneous failures the cluster survives while keeping data available. FT1 pairs with RF2 and survives one node or one disk failure. FT2 pairs with RF3, survives two, and needs at least 5 nodes. Block awareness and rack awareness extend this to whole chassis or racks when enough of them exist. Prism's Data Resiliency widget shows the current tolerance.

When a node fails, AHV HA restarts that node's HA-protected VMs on the remaining hosts. It can only do that if there is spare capacity, so a cluster at about 50% load restarts everything. The same happens if a host loses all network connectivity.

Register each Prism Element cluster with Prism Central to manage it there. You can then jump into the cluster's own Prism Element from Prism Central's cluster quick-access widget. Validate the Prism Central settings too: DNS/NTP, SMTP, its size (small, large, scale-out), and that registration is healthy.

Security options:
- RBAC (role-based access): configured in Prism Central, with roles scoped to categories or clusters.
- Directory services (AD/LDAP).
- Cluster lockdown: key-only SSH to the CVMs and hosts.
- Data-at-rest encryption.
- SCMA / STIG hardening, which self-heals drifted settings.
- The recycle bin, which keeps deleted VMs and disks for a while.

Maintenance mode is entered from Prism Element (Hardware → host → Enter Maintenance Mode) or from Prism Central. On AHV, running VMs are live-migrated off the host. VMs that can't be migrated, such as Agent VMs, GPU-passthrough VMs and host-affinity-pinned VMs, are shut down or need you to act. Exit maintenance mode when the work is done.

Expansion node types:
- HCI nodes: compute plus storage.
- Storage-only nodes: run AHV with a minimal CVM footprint, add capacity, and run no user VMs.
- Compute-only nodes: add CPU and RAM without contributing storage.

Expand Cluster in Prism Element discovers the new node, images it if needed, and adds it.

Within a node, you can add or replace disks (Hardware → Disk → Remove Disk, then insert the new one), memory and NICs, within the platform's limits. After an SSD or HDD failure, the cluster re-protects that disk's data on the other nodes.

Out-of-band management means the IPMI/BMC (iDRAC, iLO, XClarity and so on). Use it for the remote console, power control, bare-metal imaging, and the Redfish firmware updates LCM performs. Configure the OOB credentials in Prism so LCM can use them.

Upgrades go through Life Cycle Manager (LCM) in Prism Element or Prism Central. The order is: run inventory, pick the updates, pass the pre-checks, then let LCM do a rolling upgrade one node at a time. Older paths such as 1-click upgrade in Prism Element (Settings → Upgrade Software) still exist for some components.

An AOS upgrade restarts the CVMs one at a time. Guest VMs keep running on the same host while their I/O is served by the other CVMs. A hypervisor upgrade is different: it needs the host emptied, which on ESXi means DRS / live migration.

Run an LCM inventory first, every time. Inventory also updates the LCM framework itself when a newer version is available, which is how you get the latest LCM.

Connected sites: LCM downloads bundles from Nutanix over the internet. Dark sites have no internet access, so you either host the bundles on a local web server or use Direct Upload to push them straight into LCM.

Before upgrading, check the Compatibility and Interoperability Matrix on the support portal. It lists which AOS, hypervisor, Prism Central and firmware versions are supported together. Also read the upgrade paths page, because some jumps need an intermediate version.

Routine software maintenance covers the following: upgrading NCC before you run health checks, and Foundation; upgrading Prism Central; upgrading NGT in bulk across VMs; keeping AHV in step with AOS; and following the recommended upgrade order for dark sites.