Free NCP-MCI-7.5 Practice Test Questions 2026

Explanation:

The key metric here is %iowait at 83.20% — this is extremely high and dominates the CPU time breakdown.

What %iowait means:
%iowait measures the percentage of time the CPU is idle specifically while waiting for an outstanding disk I/O request to complete. A high value means the CPU has work it could be doing, but processes are blocked waiting on storage responses.

Combined with a very low %idle (6.86%) and low %user (2.32%), this output shows the CPU itself isn't the bottleneck — it's sitting around waiting for disk I/O to return.

This pattern strongly points to a storage subsystem issue: slow underlying physical drives, saturated disks, controller queuing, or a similar storage-layer bottleneck on that CVM.

Why the others are wrong:

B — Network congestion
Network congestion wouldn't show up in iostat's CPU breakdown as %iowait; that stat is specific to storage I/O, not network interrupts. There's also no %irq/%softirq data suggesting interrupt overload here.

C — CPU-bound Nutanix services
%user (2.32%) and %system (7.53%) are both low, which rules out heavy active computational/processing load. If Nutanix services were CPU-bound, you'd expect high %user or %system, not high %iowait.

D — Hypervisor CPU contention
Hypervisor CPU contention would typically show up as high %steal (time the CPU wanted to run but a hypervisor scheduled another VM instead). Here %steal is 0.00%, ruling this out entirely.

Reference:
This maps to Nutanix's guidance on using iostat on CVMs for performance triage — high %iowait is the classic signal to escalate to checking disk latency (e.g., via Prism disk I/O charts, iotop, or nodetool/curator logs) rather than assuming a compute-side issue.

Explanation:

rx_crc_errors count frames that arrive with a Cyclic Redundancy Check (CRC) that doesn't match the frame's computed checksum — meaning the frame was corrupted somewhere between the sender and the NIC's receive circuitry. This is a Layer 1 (physical layer) symptom, not a software, CPU, or configuration issue.

The most common root causes of rising CRC errors are:
A faulty, damaged, or poorly seated cable
A bad or mismatched transceiver/SFP
Electromagnetic interference (EMI) along the cable run
A failing switch port or NIC port

Why the other options are wrong:

A. Duplex mismatch — This typically produces late collisions and runt/frame errors, particularly on half-duplex links, not primarily CRC errors. Though severe mismatches can sometimes contribute to CRC issues, the classic symptom set for duplex mismatch is collisions/runts, not a large, steadily climbing rx_crc_errors counter. This isn't the "most likely" or textbook answer.

C. CPU saturation — This would show up as rx_missed_errors, rx_fifo_errors, or dropped packets due to the host not servicing the NIC ring buffer fast enough — not CRC corruption, since the frame's checksum failure happens on the wire/NIC before the CPU is involved.

D. MTU/fragmentation issues — These typically cause rx_length_errors or packet drops/fragmentation errors, not CRC failures, since MTU mismatches don't corrupt the CRC calculation itself.

Troubleshooting tip (real-world/exam relevance):
When you see climbing rx_crc_errors on an AHV host NIC, the standard remediation path is to check/replace the cable, reseat or replace the transceiver, and inspect the corresponding switch port for errors on its side as well.

Explanation:

In Nutanix AHV, capacity reservations help ensure that sufficient resources remain available for cluster resiliency and virtual machine availability.

The two relevant settings are:
Rebuild Capacity Reservation: Reserves storage capacity for rebuilding data after a disk or node failure.
Memory Reservation: Reserves memory capacity to support virtual machine availability during host failures.

To make the reserved storage and memory available for normal workloads, the administrator must configure these two reservations appropriately.

Why are the other options incorrect?

Option A
Capacity High Availability is not the relevant storage reservation setting.

Option B
Storage Capacity and Memory High Availability do not identify the two specific reservation settings.

Option C
Rebuild Capacity Reservation is relevant, but High Availability is not the specific memory reservation setting.

Option D
Correct. Identifies both the storage rebuild reservation and the memory reservation.

Exam tip:
Remember the two resource types: storage uses Rebuild Capacity Reservation, while memory uses Memory Reservation.

Explanation

Adding an SSH public key is the first step toward key-based authentication, but it does not by itself remove the risk of password-based access. To actually reduce the risk, the administrator must now disable the password authentication method entirely. This ensures that only the authorized SSH key can be used to log in, effectively eliminating brute-force or password-guessing attacks.

Here is why the other options are incorrect:

A. Enable core dumps...
This is a troubleshooting/diagnostic step, not a security hardening measure. Core dumps can potentially contain sensitive memory data and do not restrict access.

C. Enable a login banner...
A login banner is a legal notice (e.g., "Authorized access only"). While it is a good security practice for legal reasons, it does not technically prevent unauthorized access or reduce the interactive access risk itself.

D. Increase concurrent sessions...
This is unrelated to security and would likely increase the attack surface rather than reduce risk.

Explanation:

When memory overcommit is enabled, a Nutanix AHV host can allocate more virtual memory to VMs than the host physically has available.

If the VM is actively using all its allocated memory, but the guest OS pagefile usage is very low, the VM may still experience memory pressure at the hypervisor level.

The probable cause is that the hypervisor is swapping memory to disk.

Hypervisor swapping occurs when the host cannot satisfy memory demands using available physical RAM. It moves memory pages to storage, which is significantly slower than accessing RAM. This can cause high latency and poor VM performance even when the guest OS itself is not heavily using its pagefile.

Why are the other options incorrect?

A. Ballooning is not compatible with swapping.
Incorrect. AHV supports both ballooning and host swapping. Ballooning is preferred, while host swapping is used when ballooning cannot reclaim enough memory.

B. Memory ballooning is not enabled.
Incorrect. The question does not establish that ballooning is disabled. Even when ballooning is enabled, it cannot reclaim memory that the guest is actively using.

C. The hypervisor is swapping.
Correct. Host-level swapping can cause slow VM performance without high guest pagefile usage.

D. The CVM is swapping.
Incorrect. CVM swapping relates to memory pressure in the Controller VM, not directly to the guest VM's memory allocation.

Exam Tip

Remember the distinction between guest OS swapping and hypervisor swapping:

Guest OS swapping: The guest operating system moves memory pages to its own pagefile or swap space.

Hypervisor swapping: AHV moves VM memory pages to host swap storage.

Low guest pagefile usage does not rule out memory pressure at the hypervisor level.

Reference

The official Nutanix article AHV Internals: Memory Overcommit explains that AHV uses ballooning and swapping to manage overcommitted memory. It confirms that when guests actively use their memory, AHV may need to use host swapping, which can significantly affect performance.

Explanation:

Runway (Memory Runway, and similarly Storage Runway/CPU Runway) is a predictive analytics feature in Prism Central that estimates how many days remain before a resource is exhausted, based on the trend/rate of growth observed over a historical analysis window (typically looking at usage patterns over the preceding weeks).

The key insight here is that Runway is not just a snapshot of current utilization — it's a forward-looking projection. Even though current utilization is only 60% (which sounds like there's plenty of headroom), a short Runway of 12 days indicates that the rate of consumption/growth trend is steep enough that, if it continues at the same pace, the resource will be exhausted in just 12 days. This directly matches option B: memory usage has been continuously increasing over a recent period (e.g., the last 21 days), so the trendline is steep and the projected exhaustion date is soon — regardless of the fact that absolute usage is only at 60% today.

Why the other options are wrong:

A. No historical data causing a default of 12 days
Runway doesn't have a "default to 12 days" behavior when data is missing; if there's insufficient historical data, Prism typically shows Runway as unavailable/not computed, not a specific conservative number.

C. Simulated workloads scenario
This isn't a real factor affecting Runway calculations in a production cluster; Runway is based on actual observed telemetry, not "simulated" workloads (this isn't a standard Nutanix concept in this context).

D. HA in "Guaranteed Mode"
HA Guaranteed mode reserves failover capacity but doesn't directly cause a short memory Runway; it affects how much capacity is set aside for failover, not the trend-based depletion projection that drives the Runway metric.

Reference:
Prism Central's Capacity Runway/Planning feature calculates Runway using historical trend analysis (rate of change) rather than a simple current-utilization threshold, which is why a cluster can show moderate utilization (60%) but still have a short Runway if usage is trending upward quickly.

Explanation:

Nutanix AHV documentation explicitly states that if Secure Boot is enabled for a UEFI boot configuration, the IDE interface is not supported for CD-ROM or other boot-related devices. To successfully enable Secure Boot, any CD-ROM or VirtIO ISO attached to the VM must use SATA as the bus type instead of IDE.

This limitation exists because Secure Boot relies on UEFI firmware validation, and the legacy IDE controller is not compatible with the UEFI Secure Boot chain in AHV. The administrator would need to change the disk/CD-ROM bus type to SATA or SCSI before Secure Boot can be enabled.

Why the Other Options Are Not Correct

A (CVM has insufficient RAM)
The Controller VM's memory is dedicated to storage services and has no bearing on whether a guest VM can enable Secure Boot.

B (Node has insufficient cache storage)
Cache storage relates to performance tiering, not to UEFI firmware configuration capabilities.

D (VM is utilizing UEFI mode)
UEFI mode is actually a prerequisite for Secure Boot, not a blocker. Secure Boot cannot function without UEFI.

Explanation:

Foundation Central is the best tool for remotely deploying Nutanix infrastructure across multiple clusters efficiently.

It provides centralized, browser-based infrastructure imaging and deployment capabilities, allowing administrators to deploy and reimage Nutanix nodes remotely without needing to visit each cluster individually.

For a company deploying five new clusters, Foundation Central is particularly useful because it can coordinate deployments from a centralized location, reducing manual effort and improving deployment efficiency.

Why are the other options incorrect?

A. Foundation for Windows/Mac
Used to image and deploy Nutanix nodes from a Windows or Mac workstation. It is less suitable for centralized remote deployment across multiple clusters.

B. Foundation Central
Correct. Provides centralized, remote deployment and imaging capabilities.

C. Standalone Foundation VM
Used to image Nutanix nodes from a dedicated Foundation virtual machine, but it is not the best centralized deployment solution for this scenario.

D. Phoenix ISO image
Primarily used for recovery, troubleshooting, and certain node maintenance operations, not centralized deployment of multiple new clusters.

Exam Tip

Remember the distinction:

Foundation Central: Centralized, remote imaging and deployment.
Foundation for Windows/Mac: Imaging from a local workstation.
Standalone Foundation VM: Imaging from a dedicated VM.
Phoenix ISO: Recovery and troubleshooting.

Key takeaway
When a question emphasizes remotely deploying multiple Nutanix clusters efficiently, look for Foundation Central.

Explanation:

Balance-slb (Source Load Balancing): This bond mode distributes VM traffic across multiple active physical uplinks based on the source MAC address. It is the default and most popular AHV bond mode because it allows guest VMs to utilize bandwidth across multiple physical adapters without requiring any special switch-side configurations (like LACP/802.3ad) on the physical switches.

Why Other Options Are Incorrect:

A. Balance-tcp
Requires LACP (Link Aggregation Control Protocol) configuration on the upstream physical switches to bundle the ports together.

B. Active-Backup
Only uses one active physical adapter at a time for traffic forwarding, meaning it does not actively use multiple adapters simultaneously for aggregate bandwidth.

C. No Uplink Bond
Does not combine uplinks or provide a mechanism for multi-adapter load balancing.

Explanation:

The test_ahv_entering_mm_pinned_vms pre-check specifically identifies VMs that are "pinned" to a host and therefore cannot be live-migrated when the host needs to enter maintenance mode during an LCM upgrade. All three VM types listed in this scenario fall into this pinned category for well-known technical reasons:

Agent VMs — By design, Agent VMs are tied to a specific host and are not live-migrated during normal maintenance-mode operations. They typically must be powered off or Acropolis handles them specially, but they block standard live-migration workflows.

VMs with Windows Defender Credential Guard — Credential Guard relies on virtualization-based security (VBS) features that are not compatible with live migration, since the security state is tied to the specific hardware/host context.

vGPU-enabled VMs — vGPU-enabled VMs are bound to the physical GPU hardware on that specific host and cannot be live-migrated to another host unless that exact GPU profile/hardware is guaranteed elsewhere, which AHV live migration does not support for vGPU.

Since none of these VM types can be live-migrated off the host, the only way to satisfy the pre-check and allow the host to successfully enter maintenance mode is to power them off before proceeding with the LCM upgrade. Once powered off, they no longer block the host from entering maintenance mode.

Why the other options are wrong:

A. Enable DRS to migrate Credential Guard VMs
Even with DRS enabled, Credential Guard VMs still cannot be live-migrated due to the VBS/hardware-security binding. DRS doesn't override this fundamental incompatibility.

B. Update VM-VM anti-affinity rules for the Agent VM
Anti-affinity rules control VM placement relative to other VMs, not whether an Agent VM can be live-migrated. This doesn't address the root cause because Agent VMs are host-pinned by design.

C. Manually migrate the Agent VM and vGPU VM
This is technically not possible/supported. Agent VMs and vGPU-enabled VMs are pinned specifically because they cannot be migrated to another node. This option misunderstands the nature of the restriction and also omits the Credential Guard VMs entirely.

Reference:

Nutanix LCM documentation on maintenance-mode pre-checks notes that pinned VM types, including Agent VMs, GPU/vGPU VMs, and VBS/Credential Guard-enabled VMs, block automatic evacuation during host entry into maintenance mode. It recommends powering off these VMs prior to initiating the upgrade if they reside on the target host.

Explanation:

For application-level high availability (HA) of database workloads, such as Oracle RAC or Windows Failover Clustering, you need to provide shared block storage that multiple VMs can access simultaneously. On Nutanix AHV, this is achieved by creating a Volume Group and enabling shared access mode.

According to Nutanix documentation, a directly attached volume group can be configured for access by a single VM (exclusive access) or by multiple VMs (shared access). The critical caution is that shared access should only be used when the VMs are configured with cluster-aware software, which handles concurrent disk access to prevent data corruption.

Supported Products and Solutions

Oracle RAC: Bare metal and virtualized environments.
Windows Failover Clustering: Database and application HA.
Linux Guest VM Cluster (RHEL HA): High availability clustering.
VERITAS InfoScale: Storage management for HA.

For a database workload requiring application-level HA, the correct setting is to configure the Volume Group in shared mode (enabling multi-attach) so both VMs can see the same shared disks.

Why the Other Options Are Not Correct

A (External client access)
This setting is for enabling iSCSI access from external clients outside the AHV cluster, not for attaching shared storage to VMs within the same cluster.

C (vDisk on different storage container)
Storage container placement does not enable shared access. The vDisks in a Volume Group can reside on the same or different containers, but this alone does not provide the multi-attach capability required for HA clustering.

D (Multi-writer mode via storage policy)
Nutanix Volume Groups do not use a "multi-writer" storage policy setting. Shared access is configured directly on the Volume Group itself through the shared or multi-attach option, not through a separate storage policy.

Explanation:

Nutanix Volume Group (VG) vDisk load balancing (also called Volume Group Load Balancing or VGLB) distributes ownership of the vDisks in a VG across multiple Controller VMs (CVMs) for better performance on I/O-intensive workloads.

According to the AHV Administration Guide:

vDisk load balancing is disabled by default when a Volume Group is directly attached to VMs (the common AHV native attachment method).

vDisk load balancing is enabled by default when Volume Groups are attached to VMs by using a data services IP address (i.e., the iSCSI / Nutanix Volumes path that uses the cluster Data Services IP).

Therefore, to simplify management and have vDisk load balancing enabled automatically, the administrator should attach the Volume Groups using the Data Services IP address.

You can still explicitly enable load balancing on directly attached VGs with the aCLI parameter load_balance_vm_attachments=true, but that is not the “by default” method the question asks for.

Why the other options are incorrect

A – SCSI device timeout recommendations
SCSI device timeout recommendations exist for Linux VMs using load-balanced VGs, but this does not enable load balancing.

B – Direct attachment
Direct attachment actually disables load balancing by default.

C – Number of load-balanced VGs
There is a practical limit on the number of load-balanced VGs per VM (historically around 10), but this limit does not enable the feature.

References

AHV Administration Guide – Enabling Load Balancing of vDisks in a Volume Group explicitly states the default behavior for direct attachment versus Data Services IP attachment.