NCM-MCI common mistakes
NCM-MCI · Understand the traps and how to avoid them.
Treating Prism Central as a mandatory first step and sizing it like Prism Element, then being unable to answer questions about which operations are still performed in Prism Element versus Prism Central.
Candidates memorize that Prism Central enables multicloud and assume everything is centralized; they never separate cluster-local tasks (host/disk/CVM management, storage containers) from PC-scoped features (categories, policies, IAM, Flow, LCM).
What to do instead
Build a two-column map before studying: PE-only (cluster health, disk replacement, container creation, node imaging) versus PC-required (categories, protection domains via policies, RBAC, Flow, Calm, NCM). Practice stating for each task which UI/API owns it and why. Domain: Deploy and Configure Nutanix Clusters and Prism Central.
Confusing VLAN-backed subnets with overlay (VPC) networks when designing Flow or multi-tenant isolation, and misidentifying where the gateway and DHCP live in each case.
Both appear as 'networks' in Prism Central, so candidates treat them as interchangeable and pick overlay for simple L2 segregation or VLAN for routed multi-tier isolation.
What to do instead
Learn the decision rule: VLAN-backed networks extend the physical L2 and require upstream switch configuration; overlay networks are terminated by the Flow virtual switch on the AHV host and need a VPC with an external subnet for egress. For each design scenario, state the isolation boundary, the gateway location, and the DHCP source. Domain: Networking and Network Virtualization.
Describing RF2/RF3 as simple mirroring and then failing capacity questions that involve reserved capacity, container replication factor, and rebuild headroom.
Candidates learn 'RF2 means two copies' and stop there, missing that usable capacity depends on RF, container-level RF overrides, and the reserve needed to rebuild after a node loss.
What to do instead
Practice the arithmetic on a fixed example: raw capacity per node, times node count, divided by RF, minus the capacity required to sustain a single-node failure. Work through what happens when RF3 is set on a container inside an RF2 cluster. Domain: Storage Architecture, Data Protection, and Resiliency.
Treating snapshots as backups and protection domains as the only replication mechanism, without distinguishing local snapshots, async replication to a remote cluster, and Nutanix Cloud Clusters or object storage targets.
Snapshot and protection-domain terminology overlaps, so candidates collapse RPO/RTO discussions into 'take a snapshot' and misanswer DR design questions.
What to do instead
Create a matrix of snapshot only, protection policy with async replication, and backup to external target, listing RPO, RTO, failure domain covered, and the Prism Central feature that schedules each. Domain: Storage Architecture, Data Protection, and Resiliency.
Assuming roles assigned in Prism Element carry over identically to Prism Central, and ignoring the role of identity providers and categories in scoping access.
Both consoles have role-based access, so candidates answer IAM questions using PE role names and forget that PC adds AD/LDAP integration, custom roles, and category-based project scoping.
What to do instead
Memorize the PC access model: authentication source, role, and scope (category or project). Practice mapping a requirement like 'team A can only see VMs tagged project-A' to a specific role plus category scope. Domain: Security, Identity, and Access Control.
Diagnosing performance problems by looking only at VM CPU and memory graphs, without checking the CVM, storage latency, or the host-level bottleneck.
Candidates default to guest metrics because that is what they know from other hypervisors, and never build the habit of reading Prism's analysis charts across cluster, host, VM, and disk layers.
What to do instead
Adopt a fixed triage order: is it one VM, one host, or the whole cluster; then check CVM health, storage latency and IOPS, network drops, and only then guest OS. Practice reading the analysis dashboard top-down and naming the layer where the anomaly appears. Domain: Performance Monitoring, Troubleshooting, and Optimization.
Answering automation questions with generic scripting language instead of naming the specific Nutanix interfaces: Prism Central v3 APIs, v4 APIs, and the CLI tools on the CVM.
Candidates know automation conceptually but not which API version or endpoint applies to a given object, and they blur v2/v3 and v4 API semantics.
What to do instead
For each common task (create VM, list clusters, apply a category), identify the API family, the object model, and whether it is PC-scoped or PE-scoped. Note where v4 APIs differ in intent-based versus task-based behavior. Domain: Multicloud Operations, Automation, and Lifecycle Management.
Ignoring lifecycle management (LCM) and firmware/baseline concepts, then missing questions about how cluster components are updated and what prerequisites exist.
LCM is a small-sounding feature, so candidates skip it and cannot describe inventory, available updates, or why an update may be blocked.
What to do instead
Learn the LCM workflow: inventory scan, available updates, prerequisites such as connectivity to the update source and sufficient capacity, then the staged apply. Be able to explain why an update shows as unavailable. Domain: Multicloud Operations, Automation, and Lifecycle Management.
Designing a multicloud architecture without stating the failure domains, the management plane placement, and the network path between clusters.
Candidates jump to product names and features and skip the design reasoning the question is actually testing: what fails, what survives, and how the pieces connect.
What to do instead
For every design prompt, write three lines first: failure domains, where Prism Central and any shared services live, and the connectivity path (VPN, direct connect, or overlay). Then justify each component choice against those lines. Domain: Design and Plan a Nutanix Multicloud Infrastructure.
Managing VMs by memorizing menu locations rather than the underlying model: AHV as the hypervisor, the CVM as the controller VM, and how VM operations map to the API.
Candidates prepare by clicking through the UI, so scenario questions that describe an API call or a CVM-level operation leave them unable to reason about what is actually happening.
What to do instead
For each VM operation (clone, migrate, snapshot, power state change), describe what happens at the hypervisor, the CVM, and the API layer. Practice explaining why a VM cannot start in terms of host resources, storage container, and network attachment. Domain: Virtual Machine and Workload Management.