Manage Virtual Machines and Workloads
VM management items test AHV operations and the policies that keep workloads available: where a VM gets its features, and what the cluster silently reserves.
Building and protecting VMs
- A VM is defined by vCPU/RAM, disks, NICs on VLANs, and categories. Categories are the selector for every policy (protection, affinity, security, self-service).
- Nutanix Guest Tools (NGT) enables VSS-consistent snapshots, self-service restore and file-level recovery, and IP reporting. Install it on every guest you can.
- Affinity policies enforce VM-VM or VM-host placement; anti-affinity keeps cluster members apart.
- HA reservation reserves cluster capacity so that if a host fails, its VMs can restart elsewhere. Overcommitting VMs eats into that reservation.
Worked example
Two domain controllers must never run on the same host. Assign both the category role:dc, create a VM-VM anti-affinity policy on that category, and confirm the HA reservation still fits. If the reservation cannot be satisfied, additional VMs may fail to power on after a host failure - the policy is correct, but capacity is not.
Prism Central layer
Projects, self-service roles and Calm blueprints let end users deploy VMs without cluster-admin rights, while the cluster remains the execution engine.
Traps
- Taking a VM snapshot and calling it a backup - it lives in the same cluster.
- Applying a policy before the VM carries the matching category.
- Assuming a clone is space-free; clones share extents until written.
- Forgetting NGT, then discovering snapshots are crash-consistent only.