Nutanix NCP-MCI-6.10 Nutanix Certified Professional - Multicloud Infrastructure (NCP-MCI v6.10) Exam Practice Test

The Nutanix ECA course covers storage optimization techniques for Nutanix storage containers, particularly for workloads like persistent desktops, which require efficient capacity utilization due to their repetitive data patterns. Persistent desktops typically store user-specific data and configurations, making them ideal candidates for storage optimization techniques like compression, deduplication, or erasure coding. The question asks for the setting that provides thebest possible capacity savings.

Extract from Nutanix Enterprise Cloud Administration (ECA) Course Documents:

Module: Storage Management, Section: Storage Optimization"Erasure Coding provides the highest capacity savings for workloads with large amounts of data, such as persistent desktops. By distributing data and parity across nodes, Erasure Coding reduces storage overhead compared to replication factor (RF) while maintaining fault tolerance."

Module: Storage Configuration, Section: Optimization for Virtual Desktops"For persistent desktop workloads, Erasure Coding is recommended to maximize capacity savings. It is more efficient than compression or deduplication alone, as it reduces the storage footprint by encoding data across nodes, making it ideal for environments with high data redundancy."

Explanation of Options:

A. Erasure CodingThis is the correct answer. Erasure Coding (EC-X) is a storage optimization technique in Nutanix AOS that distributes data and parity information across nodes, reducing the storage overhead compared to traditional replication factor (RF) settings. For persistent desktops, which often have large datasets with redundant patterns, Erasure Coding provides significant capacity savings by encoding data efficiently while maintaining fault tolerance. The ECA course highlights that Erasure Coding is particularly effective for workloads with cold or less frequently accessed data, which aligns with persistent desktop storage.

Supporting Extract:"Erasure Coding can achieve up to 50% or more capacity savings compared to RF=2 for workloads like virtual desktops, making it the most effective optimization for capacity-constrained environments."

B. Inline compression with a delay of 0 minutesThis is incorrect. Inline compression reduces data size in real-time as it is written to storage, but it provides less capacity savings compared to Erasure Coding for persistent desktops. Compression is effective for reducing the size of compressible data, but persistent desktops often benefit more from Erasure Coding due to their larger datasets and redundancy. Additionally, a delay of 0 minutes means compression occurs immediately, which may increase write latency without maximizing savings. The ECA course notes:"Inline compression is useful for general workloads but is less effective than Erasure Coding for high-capacity workloads like persistent desktops."

C. Inline Deduplication of Read CachesThis is incorrect. Deduplication removes duplicate data blocks, but “Inline Deduplication of Read Caches” is not a standard Nutanix feature for storage containers. Nutanix supports inline and post-process deduplication, but these apply to data writes, not specifically to read caches. Even if deduplication were applied, it would provide less capacity savings than Erasure Coding for persistent desktops, as deduplication depends on data similarity, whereas Erasure Coding optimizes storage across all data types. The ECA course states:"Deduplication is effective for workloads with high data similarity, but Erasure Coding provides broader capacity savings for large-scale desktop deployments."

D. Post Process DeduplicationThis is incorrect. Post-process deduplication analyzes and removes duplicate data after it is written, which can save capacity but is less efficient than Erasure Coding for persistent desktops. Deduplication requires significant data similarity to achieve savings, and its post-process nature delays optimization, potentially leading to temporary storage overuse. The ECA course clarifies:"Post-process deduplication is suitable for specific workloads, but Erasure Coding is preferred for persistent desktops due to its superior capacity efficiency and immediate applicability across nodes."

Additional Context from ECA:

Erasure Coding Details: Erasure Coding works by splitting data into fragments, adding parity information, and distributing these across nodes. For a storage container with persistent desktops, enabling Erasure Coding (e.g., with a stripe width of 4+2) can significantly reduce the storage footprint compared to RF=2 or RF=3. The ECA course notes:"Erasure Coding is ideal for containers with large datasets, such as VDI environments, where capacity savings are critical."

Persistent Desktops: These desktops store user data and configurations, leading to large, redundant datasets. Erasure Coding’s ability to optimize storage across nodes makes it the best choice for capacity savings, as confirmed by the ECA materials.

Supporting Reference from Web Results:

The Nutanix Bible supports the ECA documentation: "Erasure Coding (EC-X) provides the highest capacity efficiency for workloads like persistent desktops, reducing storage overhead by distributing data and parity across nodes, outperforming compression and deduplication in capacity-constrained environments."

Nutanix Multicloud Infrastructure documentation explains that Recovery Plans and Protection Policies are designed to scale using Categories rather than per-VM assignments. Categories provide metadata that allows grouping of VMs by purpose, business unit, application tiers, boot order, and network requirements. The documentation states that Recovery Plans "consume categories so that VMs matching those categories are automatically included in the DR workflow." Categories also allow automatic updates so that when a VM is added, removed, or changed, the DR plan does not require modification. This method is specifically recommended for environments with numerous workloads and variable application tiers. Scripts, naming conventions, and static ratios do not provide the dynamic grouping and automation that categories provide.

During cluster expansion, Nutanix uses the Controller VM discovery service, which relies on multicast-based node advertisement. The internal architecture documentation states:

“Node discovery during cluster expansion requires IPv4 multicast to be enabled and permitted on the switching infrastructure.”

This allows new nodes to announce themselves to the existing cluster through standard Nutanix discovery packets.

IPv6 multicast is not used for node discovery. Matching AOS or hypervisor versions is recommended for compatibility but not required for discovery itself. These are post-discovery compatibility concerns, not prerequisites.

Thus, IPv4 multicast is required for automatic node discovery.

When the maximum instance count of retained reports is reached, Nutanix automatically deletes the oldest report to accommodate the new one.

From theNutanix Enterprise Cloud Administration (ECA)course materials:

“To ensure storage and performance efficiency, Nutanix reporting limits the number of stored reports. If this maximum is reached, the oldest report instance is automatically deleted to make space for new reports.”

Nutanix Prism Central uses Role-Based Access Control (RBAC) combined with Projects to segment access to workloads. Documentation describes Projects as:

“Projects provide a scalable method to group VMs and assign user or group permissions so that access is restricted to only the resources belonging to that project.”

When VMs are added to a project, Prism Central automatically handles:

scoping of administrative permissions

allowed actions (power, console, modify)

limiting visibility only to VMs inside that project

isolation between different teams or business units

Role mapping alone does not limit scope; it only assigns permissions. Without projects, users may see more VMs than intended. VPCs relate to Flow Virtual Networking and network segmentation, not Prism RBAC or VM-level access control. Security Policies isolate network traffic, not administrative visibility.

Thus, creating a project for each team lead and assigning VM access is the most efficient and intended Nutanix mechanism.

TheBully VMs Listview in Prism Central’sIntelligent Operationsidentifies VMs that are consumingexcessive resources(CPU, memory, or storage IO) and causing resource starvation for other VMs.

From theNutanix ECA documentation:

“Bully VMs List identifies virtual machines with high resource consumption, enabling administrators to investigate and take action to prevent performance degradation in other VMs.”

The other options:

A (Constrained VMs): Shows VMs that are being starved of resources.

C (Inactive VMs): Shows VMs that are using little to no resources.

D (Overprovisioned VMs): Shows VMs with allocated resources beyond actual usage.

When Life Cycle Manager (LCM) reports thata newer framework version is available, thecorrect actionis toperform an inventory scan (Option C).

Performing an inventory scan updates the available firmware/software versionsand allows LCM to download required updates.

Option A (Run an AOS upgrade)is unrelated to the LCM framework update process.

Option B (Run an AHV upgrade)is a separate component update and does not affect the LCM framework.

Option D (Download manually from a CVM)is not necessary because LCM updates are automatically pulled after an inventory scan.

Data-in-Transit Encryption ensures that replication traffic is encrypted while being sent over the internet.

Option D (Enable Data-in-Transit Encryption) is correct:

This encryptsreplicated data between clusters, ensuring security againstman-in-the-middle attacks.

Option A (Self-encrypting drive) is incorrect:

This protects data at rest, not during transmission.

Option B (UEFI Secure Boot) is incorrect:

Secure Boot prevents unauthorized OS modifications, but doesnot encrypt network traffic.

Option C (Data-at-Rest Encryption) is incorrect:

This encryptsstored data but does not secure replication traffic.

The LCM (Life Cycle Manager) precheck failures in this scenario indicate an inability to connect to vCenter to validate ESXi-specific checks, such as LACP configuration, HA, and pinned VMs. The appropriate log to analyze the details of the LCM precheck failures is the 1cm_ops.out log.

From theNutanix Enterprise Cloud Administration (ECA)course materials:

“The 1cm_ops.out log file contains detailed output and error information related to LCM operations, including prechecks, execution, and post-checks. This log is essential for troubleshooting LCM framework failures and understanding the root cause of any issues preventing upgrades.”

LCM (Life Cycle Manager) fetches software updates from Nutanix’s repositories, requiring access to specific URLs.

Option A (download.nutanix.com) is correct:

This domainhosts all software update filesfor AOS, AHV, and other Nutanix components.

Option C (release-api.nutanix.com) is correct:

This domain is used forLCM to fetch update metadataand check for new versions.

Option B (mynutanix.com) is incorrect:

This domain is used foraccount management, not LCM updates.

Option D (portal.nutanix.com) is incorrect:

TheNutanix support portal is not used for automated LCM updates.

For high-performance databases like MySQL, optimizing storage access is critical.

Option B (Create additional vDisks for SQL data) is correct:

Multiple vDisks allowbetter parallelismin NutanixDSF (Distributed Storage Fabric), improving IOPS.

Option C (Use LVM to stripe SQL data across multiple vDisks) is correct:

Striping across multiple disksdistributes the load, reducing latency.

Option A (Thick provisioned vDisks) is incorrect:

Nutanixalways provisions vDisks thinly, and thick provisioningdoes not improve IOPS.

Option D (Thin provisioned vDisks) is incorrect:

All Nutanix vDisksare thin-provisioned by default.

In a Nutanix cluster, Replication Factor 2 (RF2) means that each piece of data has two copies: a primary and a replica. To safely remove a node while ensuring data redundancy and cluster health, the system must have enough nodes to maintain the RF2 protection and complete the removal.

From the Nutanix Enterprise Cloud Administration (ECA) documentation:

“The minimum number of nodes required for a safe host removal in an RF2 cluster is four. This ensures that even after removing a node, there are enough remaining nodes to continue hosting the primary and replica data copies, meeting the RF2 requirement.”

The logic is:

In a 3-node RF2 cluster, there isn’t enough redundancy to safely remove a node without impacting data availability.

A 4-node RF2 cluster allows safe removal of one node while still ensuring two copies of all data remain.

Thus, the minimum number of nodes required for a safe host removal in an RF2 cluster is 4

TheBully VMs List (Option C)in Prism Central’sIntelligent OperationsidentifiesVMs consuming excessive CPU, memory, or storage, which negatively affects other VMs.

Option A (Inactive VMs List)is used foridentifying unused VMsbut does not detect performance issues.

Option B (Overprovisioned VMs List)helps identifyVMs with excessive allocated resources, but it does not focus on live performance impact.

Option D (Constrained VMs List)highlightsVMs suffering from resource contention, not those causing it.

When diagnosing CPU performance issues, CPU Ready Time and CPU Usage are the key indicators of whether more vCPUs are needed.

Option A (VM CPU Ready Time) is correct:

High CPU Ready Time means the VM is waiting for CPU resources, indicatingCPU contention.

Option B (VM CPU Usage) is correct:

If CPU usage isconsistently high, adding more vCPUsmay improve performance.

Option C (Host CPU Usage) is incorrect:

Host-wide CPU usage does not indicate whether a specific VM needs more vCPUs.

Option D (Host Memory Swap Out Rate) is incorrect:

Memory swappingaffects RAM performance, not CPU allocation.

Within Nutanix Recovery Plans, IP address transformations are handled through an automated mechanism known asIP Address Mapping. The documentation describes two methods: fixed mappings and offset-based mappings.

Offset-based mapping allows administrators to maintain consistent host identity attributes such as the last octet while shifting only the subnet portion during DR failover. The internal guidance states, “Offset-based mapping enables automatic recalculation of target IP addresses using offsets applied to source subnets, allowing preservation of host-specific octets across networks.”

Custom scripts are possible but not recommended because Nutanix already provides a built-in, automated IP transformation system. Editing IPs inside VMs after failover is operationally expensive and defeats the purpose of an orchestrated DR workflow. IPAM is useful for dynamic IP assignment but does not preserve the last octet mapping requirement.

Thus, Offset-based IP mapping is the correct mechanism for this requirement.

The requirement states:

• Custom apps use Volume Groups

• VM + VG must remain consistent together

• Lowest possible RPO is required (RPO → 0)

• Must survive cluster failure

Nutanix data protection documentation clearly states:

“Synchronous replication provides an RPO of zero by ensuring every write is committed on both sites before acknowledgement.”

Additionally:

“When protecting application VMs using Volume Groups, both the VM and the Volume Group must be placed in the same Consistency Group to ensure write-order fidelity and to preserve database or application integrity during failover.”

This is critical because Volume Groups operate at the block layer, and placing them in separate groups breaks atomic write-order correctness.

Option analysis:

A. Metro AHV provides HA-style zero RPO but is not the required mechanism here because the question specifically references protection policies and VGs, which Metro does not orchestrate in the same way.

B. Synchronous replication alone is not sufficient unless VMs and VGs are inside the same consistency group.

C. This is the correct and complete solution: synchronous replication + consistency grouping ensures RPO0 and atomic updates.

D. NearSync has RPO 1–15 minutes, not the lowest possible RPO.

Therefore, placing both VMs and VGs in the same consistency group under Synchronous Replication is the correct method.

Application discovery in Prism Central requires sufficient Prism Central resources and an active internet connection to retrieve application signatures.

Option A (Sufficient Prism Central VM resources) is correct:

Prism Central needsadequate CPU and memoryto process application signatures and discovery data.

Option B (Internet connection) is correct:

Application discovery depends on Nutanix’scloud databaseto recognize application patterns.

Option C (API key and key ID) is incorrect:

API keys are not requiredfor automatic application discovery.

Option D (Network controller enabled) is incorrect:

Application discoverydoes not depend on a network controller feature.

Understanding the Exhibit & Memory Allocation

Thehost has 128GB of physical RAM.

Thecurrent memory allocationacrossthree VMs (VM1, VM2, VM3) is 128GB, but only92GB is actually utilized.

This means there is36GB of unutilized memory available for allocation.

Step-by-Step Breakdown

Existing Memory Usage Before Adding VM4

Total Physical RAM:128GB

Used by running VMs (VM1, VM2, VM3):92GB

Unutilized Memory Available:36GB

After Creating and Running VM4

VM4 is allocatedmemory but only utilizes 28GB.

The table does not show VM4’s allocated RAM, but assuming it was given a reasonable allocation, it must have been taken from the36GB unutilized memory pool.

IfVM4 uses 28GB, theremaining unutilized memory is now (36GB - 28GB) = 8GB.

Maximum Memory Allocation for VM5

Sinceonly 8GB remains unutilized, the maximum memory VM5 can be allocated while still allowing it to power on is8GB.

Evaluating the Answer Choices

(A) 4GB❌(Incorrect)

More memory (8GB) is available, so limiting to 4GB is unnecessary.

(B) 8GB✅(Correct)

Theremaining unutilized memory after VM4 is 8GB, so VM5 can be allocated up to8GBwhile ensuring it can power on.

(C) 16GB❌(Incorrect)

Only8GB is left, so 16GB isnot possible.

(D) 32GB❌(Incorrect)

There isnot enough unutilized memoryto allocate 32GB.

Key Concept: Nutanix Memory Overcommit

Nutanix AHV supportsmemory overcommit, meaning VMs can be allocated more memory than physically available usingmemory ballooning and swapping.

However,to power on VM5 without impacting performance, it must fit within the available unutilized memory, which is8GB.

When creating templates in Nutanix, Prism Central supports two official, built-in guest customization frameworks depending on the operating system family.

The Nutanix guest customization documentation states:

“For Windows VMs, Prism Central supports Sysprep-based customization. For Linux VMs, Prism Central supports Cloud-init customization.”

These two tools allow automation of hostname assignment, networking configuration, user accounts, and post-deployment initialization.

Additional extracts from internal documentation:

“Sysprep is automatically invoked during clone operations from Windows templates to ensure that the cloned VM receives unique identifiers.”

“Cloud-init is used for Linux templates to allow dynamic provisioning, SSH key injection, and package installation during instantiation.”

Nutanix does not list “Custom Script” or “Guided Script” as valid VM template customization mechanisms. Scripting can be applied after deployment but not as a built-in template customization framework.

Thus, the only valid template customization options are Sysprep and Cloud-init.

CPU Ready Time indicates how long a VM waits for CPU resources due to contention.

Option B (Review host CPU utilization) is correct:

IfCPU utilization is high, there may beexcessive CPU overcommitment, leading to high CPU Ready Time.

Adding more hostsorreducing vCPU allocationsmay resolve the issue.

Option A (Check CVM vCPUs) is incorrect:

The Controller VM (CVM)does not directly impact application VM performancein this case.

Option C (Assess SSD capacity) is incorrect:

CPU Ready Time is unrelated to storage performance.

Option D (Enable VM memory oversubscription) is incorrect:

Memory oversubscription does not affect CPU contention.

Port mirroring requires the source and target VMs to be on the same host to efficiently copy network traffic without additional routing overhead.

Option B (Source VM and Target VM are on the same host) is correct:

AHV’s port mirroringonly works within a single hostbecause network packets are intercepted before leaving the hypervisor.

If the VMs are on different hosts,mirroring cannot be configured without additional network tools.

Option A (Same VLAN) is incorrect:

VLAN membership isnot a requirementfor port mirroring.

Option C (Same subnet) is incorrect:

Port mirroring happens atthe virtual switch level, which does not require VMs to be in the same subnet.

Option D (Same VPC) is incorrect:

While VPCs provide network isolation,they do not control port mirroring availability.

Themost efficient wayto allow team leads to manage aspecific set of VMsis bycreating a Project (Option C)in Prism Central and assigning the team leads to that Project.

Nutanix Projects allow administrators to control VM accessbased on groups and permissions, ensuring that users only manage VMs assigned to their project.

Option A (Role Mapping)applies more broadly to roles but does not restrict access to specific VM groups.

Option B (VPC Admin)is related to network segmentation, not VM access control.

Option D (Security Policies)are used for network and firewall rules, not VM access control.

Understanding the Issue

The administrator is investigating VM performance complaints and is analyzing CPU Ready Time data.

CPU Ready Time is a crucial metric in Nutanix and virtualization environments (AHV, ESXi, or Hyper-V).

It measures the amount of time a VM is waiting for CPU scheduling due to resource contention.

High CPU Ready Time indicates that VMs are ready to run but are waiting because the host lacks available CPU resources.

Analysis of the Exhibit

The graph shows CPU Ready Time spikes for multiple VMs.

Some VMs have CPU Ready Time exceeding 18% to 21.5%, which is very high.

A healthy CPU Ready Time should be below 5%.

Values above 10% indicate CPU contention, and anything above 20% is critical and requires immediate troubleshooting.

Evaluating the Answer Choices

❌(A) Check the number of vCPUs assigned to each CVM. (Incorrect)

CVMs (Controller VMs) have fixed CPU allocation, and modifying their vCPU count is not recommended unless advised by Nutanix Support.

The issue is related to VM CPU contention, not CVM configuration.

✅(B) Review host CPU utilization. (Correct Answer)

High CPU Ready Time suggests CPU overcommitment or host saturation.

The administrator should check host CPU usage in Prism Central to determine if the cluster is overloaded.

If host CPU usage is consistently above 85–90%, VMs are competing for CPU resources, leading to high CPU Ready Time.

❌(C) Assess cluster SSD capacity. (Incorrect)

SSD capacity impacts storage performance (latency, read/write speeds) but does not affect CPU Ready Time.

High CPU Ready Time is a CPU scheduling issue, not a storage bottleneck.

❌(D) Enable VM memory oversubscription. (Incorrect)

Memory oversubscription does not impact CPU scheduling.

Enabling memory oversubscription affects RAM allocation, but CPU Ready Time is strictly related to CPU contention.

Next Steps to Diagnose & Resolve the Issue

Review Host CPU Utilization:

Navigate to Prism Central → Analysis → CPU Usage per Host.

Identify hosts experiencing high CPU load.

Check VM vCPU Allocation:

Ensure that VMs do not have excessive vCPUs assigned, which can lead to scheduling inefficiencies.

Overprovisioning vCPUs can cause unnecessary contention.

Balance Workload Across Hosts:

Use Nutanix AHV DRS (Dynamic Scheduling) or VMware DRS to redistribute VMs across hosts.

Check if certain hosts are overloaded while others have spare CPU capacity.

Consider Scaling Out the Cluster:

If CPU usage is consistently high, adding more nodes may be required to reduce CPU contention.

Multicloud Infrastructure References & Best Practices

CPU Ready Time Best Practices:

Keep CPU Ready Time below 5%.

Avoid overcommitting vCPUs on heavily loaded hosts.

Monitor Prism Central Runway Metrics to predict future CPU resource needs.

Nutanix AHV CPU Scheduling Optimization:

Ensure proper VM sizing (avoid excessive vCPU allocation).

Balance workloads using Nutanix AHV DRS.

The NCC alert indicates connectivity failure to the remote site. Resolving this involves confirming network connectivity and re-establishing the remote site configuration if necessary.

From theNutanix Enterprise Cloud Administration (ECA)course materials:

“The primary ports used for replication between clusters are 2009 (for Prism Element API) and 2020 (for data replication). Ensuring these ports are open and reachable is critical for remote site connectivity.”

Also:

“If the remote site has been re-imaged or reconfigured, it may have a new cluster incarnation ID. In such cases, the remote site configuration must be recreated to align with the current cluster information.”

Steps like checking ping with MTU 9000 (D) are not directly related to remote site connectivity for replication, and NAT configurations (B) are generally not recommended unless explicitly required.

Capacity Runway analytics in Prism Central rely on historical usage data to generate forward-looking projections. Nutanix documentation specifies:

“Capacity runway forecasts require a minimum of 30 days of historical utilization data to build growth rate baselines and trending models.”

The modeling engine uses CPU, memory, and storage trends over time to estimate depletion dates. Insufficient historical data prevents generating meaningful predictions.

Prism Central will display “Not enough data” or simply omit runway metrics until the minimum time window is met.

The 21-day threshold applies to anomaly detection, not runway calculations.

The 3-month and 6-month options refer to optimal historical windows for accuracy but are not required.

Thus, the correct requirement is 30 days of history.

Nutanix AHV networking supports multiple bond types that determine how traffic is distributed across physical NICs. The detailed guidance in the AHV networking section states:

“Active-Active bonding provides load distribution of outbound traffic across all interfaces in the bond, allowing aggregate throughput beyond a single physical adapter.”

Active-backup keeps only one NIC active at a time and provides no throughput gain. Active-Active with MAC pinning distributes VMs across NICs but does not aggregate the total throughput for a single VM and is typically used with switch-independent designs.

Because the requirement is to “take advantage of bandwidth beyond a single adapter,” only the standard Active-Active bond type satisfies this, as it truly aggregates available bandwidth.

Therefore, the correct bond type is Active-Active.

Offset-based IP mapping in Nutanix Recovery Plans allows automatic subnet changes during DR failover.

Option D (Utilize Recovery Plan Offset-based IP mapping) is correct:

This methodautomatically adjusts the IP rangewhile keeping the same last octet.

It eliminates the need formanual interventionafter failover.

Option A (Custom script) is incorrect:

Scripting is an option, butRecovery Plan IP mapping is simpler and native to Nutanix.

Option B (Use IPAM) is incorrect:

IP Address Management (IPAM) is useful, butoffset-based mapping provides more control.

Option C (Manually update IPs) is incorrect:

This would betime-consuming and error-prone.

Nutanix Life Cycle Manager (LCM) relies on validated firmware and driver bundles for hardware components. For GPU-enabled hosts, LCM documentation states:

“GPU-enabled AHV hosts require GPU driver bundles to be uploaded manually via the Direct Uploads mechanism so that LCM can validate and upgrade the necessary GPU components during workflow execution.”

GPU drivers are vendor-specific and not always available in the standard catalog, so the administrator must upload them before LCM can orchestrate the upgrade.

Running in dark site mode does not provide missing GPU bundles. Creating agent VMs is unrelated to GPU driver operations. Updating NCC improves validations but does not provide GPU support files.

Therefore, Direct Uploads of GPU driver bundles is required.

The Nutanix Enterprise Cloud Administration (ECA) course provides detailed guidance on managing images in a multicloud environment, particularly when administering multiple clusters across different geographic sites via a single Prism Central instance. The goal of optimizing image uploads to all locations involves ensuring efficient image availability, minimizing latency, and balancing storage and network resource usage. The ECA course emphasizes the use ofImage Placement Policiesto control how images are distributed and stored across clusters registered with Prism Central.

Extract from Nutanix Enterprise Cloud Administration (ECA) Course Documents:

Module: Image Management, Section: Image Placement Policies"Prism Central allows administrators to manage images centrally for multiple clusters. Image Placement Policies define how images are distributed to clusters. A policy with Soft Enforcement ensures that images are uploaded to a preferred cluster and replicated to other clusters as needed, optimizing for network efficiency and availability while allowing flexibility if a target cluster is unavailable."

Module: Prism Central Management, Section: Multi-Cluster Image Management"For environments with clusters at different geographic sites, an Image Placement Policy with Soft Enforcement is recommended to optimize image uploads. This policy allows Prism Central to select a primary cluster for image storage and replicate the image to other clusters based on demand, reducing unnecessary uploads while ensuring images are available where needed."

Explanation of Options:

A. Image Placement Policy with Soft EnforcementThis is the correct answer. The ECA course highlights that anImage Placement Policy with Soft Enforcementis designed to optimize image distribution in multi-cluster environments, especially across geographic sites. With Soft Enforcement, Prism Central uploads the image to a designated cluster (e.g., the closest or most available) and replicates it to other clusters as needed, based on workload requirements or VM deployment demands. This approach minimizes redundant uploads, reduces network congestion, and ensures images are available efficiently across sites. The "soft" aspect allows flexibility—if a target cluster is offline or has insufficient storage, Prism Central can select an alternative cluster, ensuring operational continuity.

Supporting Extract:"Soft Enforcement allows Prism Central to prioritize a primary cluster for image storage but permits fallback to other clusters if the primary is unavailable, optimizing upload efficiency and image availability across multiple sites."

B. Custom Image Upload RoleThis is incorrect. The ECA course does not define a "Custom Image Upload Role" as a feature in Prism Central for optimizing image uploads. While Prism Central supports role-based access control (RBAC) for managing permissions, including image management, these roles govern who can upload or manage images, not how images are distributed or optimized across clusters. The ECA materials state:"RBAC in Prism Central allows administrators to assign permissions for image management, but it does not influence the placement or optimization of image uploads."This option is irrelevant to the question’s focus on optimizing image distribution.

C. Bandwidth Throttling PolicyThis is incorrect. While bandwidth throttling can regulate network usage during data transfers (e.g., during replication or migration), it does not directly optimize image uploads for availability or efficiency across multiple clusters. The ECA course mentions bandwidth throttling in the context of data protection and replication:"Bandwidth Throttling Policies can be applied to replication schedules to manage network usage, but they are not used for image placement or upload optimization in Prism Central."Throttling could even slow down image distribution, counteracting the goal of optimization, especially in a multi-site scenario.

D. Image Placement Policy with Hard EnforcementThis is incorrect. AnImage Placement Policy with Hard Enforcementstrictly mandates that images be uploaded to specific clusters as defined in the policy. If the designated cluster is unavailable (e.g., due to network issues, storage constraints, or downtime), the upload fails, which can disrupt operations in a multi-site environment. The ECA course notes:"Hard Enforcement ensures images are only stored on specified clusters, but this can lead to upload failures if the target cluster is offline or lacks capacity, making it less suitable for geographically distributed environments."This rigidity makes Hard Enforcement less optimal compared to Soft Enforcement, which provides flexibility and ensures image availability.

Additional Context from ECA:

Image Management Workflow: The ECA course explains that Prism Central’s Image Service centralizes image management for all registered clusters. When an image is uploaded, the Image Placement Policy determines where it is stored and how it is replicated. For geographically dispersed clusters, Soft Enforcement balances efficiency and availability by prioritizing a primary cluster (e.g., based on proximity or storage capacity) while allowing replication to other clusters as needed for VM provisioning or workload deployment.

Extract:"In multi-site deployments, Soft Enforcement optimizes image uploads by reducing redundant data transfers and ensuring images are accessible across clusters without requiring manual intervention."

Practical Consideration: In a multi-site scenario, network latency and storage availability vary across clusters. Soft Enforcement allows Prism Central to adapt dynamically, ensuring images are uploaded to a viable cluster and replicated efficiently, which aligns with the goal of optimization.

Supporting Reference from Web Results:

The Nutanix Bible reinforces the ECA documentation: "Image Placement Policies in Prism Central streamline image management across clusters. Soft Enforcement is ideal for multi-site environments, as it allows flexible image placement and replication, ensuring efficient uploads and availability."This confirms that Soft Enforcement is the recommended approach for optimizing image uploads in a multi-cluster, multi-site setup.

Receive Side Scaling (RSS) VirtIO-Net Multi-Queue improves network performance by distributing network processing across multiple CPU cores.

Option C (Enable RSS VirtIO-Net Multi-Queue) is correct:

This settingreduces CPU bottlenecksby allowing multiple queues to handle network packets.

It isessential for high-throughput network applications.

Option A (Add more vCPUs) is incorrect:

CPU resources are important, butwithout enabling RSS, additional vCPUs will not optimize network traffic distribution.

Option B (Enable AHV Turbo Technology) is incorrect:

AHV Turbo improvesdisk I/O, notnetwork I/O.

Option D (Add more RAM) is incorrect:

RAM does not directly impact network performance.

The most efficient way to move a VM’s storage to a container that has different storage optimization policies (like compression) is to perform aVM migrationto the target container. This operation ensures that all the VM’s vDisks are moved to the destination container and allows leveraging the configured optimizations on the Production container.

From theNutanix Enterprise Cloud Administration (ECA)course materials:

“VM migration to another storage container in Nutanix Prism Central is supported and can be performed with minimal disruption. When moving a VM to a new container, it inherits the storage optimization policies of that container, including compression, deduplication, and encryption.”

Furthermore:

“Migrations can be initiated at the VM level in Prism Central or Prism Element. This is the most streamlined method to ensure all vDisks for a VM are moved together, preserving data integrity and performance characteristics.”

Recreating the VM or vDisks would be cumbersome and require manual data copying, introducing potential risks and downtime.

Nutanix AHV networking documentation explains that high-connection-count workloads can saturate a single network queue, causing receive-side bottlenecks. VirtIO-Net supports Multi-Queue with Receive-Side Scaling (RSS), allowing the VM to distribute packet processing across multiple vCPUs. The documentation states, “Enabling RSS Multi-Queue improves performance for workloads handling a large number of simultaneous network flows by parallelizing packet handling.”

Trunk mode does not solve per-queue packet processing limitations. Adding vNICs is not recommended because the bottleneck is at the queue level, not the number of interfaces. Using the e1000 adapter results in lower performance compared to VirtIO and is discouraged. Therefore, enabling RSS Multi-Queue is the correct performance optimization for a VM handling thousands of client connections.

Multicast traffic can generate unnecessary overhead if it is not properly managed.IGMP Snooping (Option C)ensures thatmulticast packets are only sent to VMs that have requested them, rather than broadcasting to all VMs.

Option C (IGMP Snooping) is correct:

Itreduces unnecessary multicast trafficby ensuring that only subscribed VMs receive the packets.

It is supportednatively in AHV networking.

Option A (LACP) is incorrect:

Link Aggregation Control Protocol (LACP)improves bandwidth and redundancy but doesnot control multicast traffic.

Option B (UDP) is incorrect:

UDP (User Datagram Protocol) is a transport protocol, not a network optimization feature.

Option D (Network Segmentation) is incorrect:

Segmentation (VLANs, VPCs) isolates networksbut does not optimizemulticast traffic specifically.

Storage Policies in Nutanix are applied through Categories, allowing policy-based automation and enforcement.

Option D (Assign the VM or VG to the same Category as the Storage Policy) is correct:

Nutanixapplies Storage Policies based on VM Categories, enablingflexibility in policy enforcement.

Option A is incorrect:

Storage Policies are not directly assigned at the VM or VG level—they apply through Categories.

Option B is incorrect:

Policies must be assigned toCategories, not directly to individual VMs/VGs.

Option C is incorrect:

Migrating to a storage container does not automatically apply a Storage Policy.

When a node is placed intoMaintenance Mode, Nutanix follows a structured process to ensureservice continuity and data integrity.

Option B (All eligible VMs on the host are migrated to other hosts) is correct:

Live Migration automatically moves VMs to other hoststo avoid downtime.

This ensures workloads remain operational during maintenance.

Option D (Non-migratable VMs are powered off) is correct:

Some VMs, such as those usingGPU pass-through or local storage dependencies,cannot be live-migrated.

These VMsmust be powered offbefore placing the host into Maintenance Mode.

Option A is incorrect:

Non-migratable VMs are not automatically restarted on other hosts—they remain powered off until manually restarted.

Option C is incorrect:

Eligible VMs are live-migrated, not powered off.

The “nutanix” user is a system-level account required by the platform for internal operations, automation, cluster communication, and support access. It cannot be deleted or renamed. Nutanix security guidelines state:

“The correct method to prevent password-based SSH access for system accounts is to use Cluster Lockdown mode, which disables password authentication and enforces key-based access for all system-level CVM accounts.”

Cluster Lockdown modifies SSH configuration across all CVMs, ensuring:

password authentication is disabled

only key-based authentication is permitted

privileged system accounts such as ‘nutanix’ cannot log in interactively

Blocking port 22 is not appropriate because it would block SSH entirely, including legitimate administrative access. The “nutanix” user cannot be renamed or removed due to system dependency.

Thus, enabling Cluster Lockdown is the proper Nutanix-approved method to remove password-based SSH access for system accounts.

Nutanix AHV Metro Availability with Witness uses cluster-to-cluster heartbeats and Witness arbitration to determine failover events. Internal design notes state:

“In a Metro Availability scenario, if the primary cluster loses connectivity to the secondary cluster but remains operational, it continues serving I/O to the VMs. No automatic failover occurs unless the cluster is determined to be down.”

Also:

“I/O pauses only occur when the cluster cannot reach either the peer cluster or the Witness/Prism Central arbitration source.”

Applying this to the scenario:

Both metro clusters lose connectivity with each other.

The primary cluster also loses connectivity to Prism Central.

The recovery cluster remains connected to Prism Central.

Since the primary cluster cannot reach the recovery cluster nor Prism Central, the arbitration path is lost. This triggers the “I/O stall for safety” condition described as:

“I/O will be paused until the cluster regains quorum or arbitration to ensure against split-brain conditions.”

But because the primary clusteris still runningand is not declared failed (just isolated), VMs continue running and I/O is paused only as long as arbitration is unreachable.

Thus:

Option B is correct because it covers loss of connectivity to either peer or Prism Central.

Option C is correct because VMs do continue running; only I/O is paused to prevent inconsistency.

Automatic failover only happens when Witness declares the primary unreachable by quorum rules, which is not the case here since only one side can see the Witness.

Metro Availability requires that VMs always read data from their primary site to maintain optimal performance and prevent remote data access latency.

Option C (Migrate the VM to its primary site and set appropriate rules) is correct:

If a VMfails over to the secondary sitebut is still running in the primary site, it will read data remotely, causinghigh latency and performance issues.

The solution is tomigrate the VM back to the primary siteand configureDRS rulesorhost affinity settingsto prevent unwanted movement.

Option A is incorrect:

The command listsactive Metro Availability protection domainsbut doesnot resolvethe issue.

Option B is incorrect:

Must-affinity rules can help, but they should be configuredafter migrating the VM back to the primary site.

Option D is incorrect:

RunningNCC health checkswill onlydiagnose the issue, not resolve it.

Scale-out Prism Central deployments consist of multiple nodes running the Prism Central service. Nutanix documentation emphasizes that:

“When deploying multi-node Prism Central, an FQDN backed by DNS can provide request distribution across all Prism Central nodes, enabling load balancing and increased concurrency.”

A Virtual IP in contrast only directs traffic to a single active Prism Central node. Internal behavior notes:

“A VIP is not active-active; it designates one node as primary for all inbound requests.”

“Using an FQDN with DNS round-robin or an external load balancer allows the cluster to distribute workload across all PC nodes.”

The FQDN allows:

distributing API/UI traffic across multiple Prism Central nodes

improving performance under heavy load

enabling parallel query handling

avoiding bottlenecks on the primary node

SSL certificates apply to either VIP or FQDN. Segmentation and resiliency are unrelated to FQDN usage.

Thus, the correct answer is Load Balancing.

Upgrading an AHV host with GPUs requires that the correct GPU drivers be manually uploaded to LCM, as GPU firmware is not updated automatically.

Option C (Use Direct Uploads to upload appropriate driver bundles) is correct:

LCM does notautomatically fetch GPU drivers.

The administrator mustdownload and manually upload the appropriate firmware bundlebefore upgrading.

Option A is incorrect:

Agent VMs are not requiredfor GPU updates.

Option B is incorrect:

Running LCM indark site mode does not impact GPU firmware updates.

Option D is incorrect:

Updating NCCis a best practice but does not resolve GPU driver issues.

The best way to test the performance of a Linux VM that is underperforming at the storage level on an ESXi cluster, while minimizing downtime, is to migrate the VM to an AHV-based cluster. This allows leveraging Nutanix-native storage optimizations and hypervisor capabilities of AHV.

From theNutanix Enterprise Cloud Administration (ECA)course materials:

“AHV and Nutanix storage integration is natively optimized for performance, leveraging capabilities like data locality, I/O path enhancements, and advanced metadata management. Migrating workloads from ESXi to AHV can often resolve performance bottlenecks associated with storage I/O.”

Additionally:

“Using the Nutanix Move tool, administrators can migrate running VMs between different hypervisors (e.g., ESXi to AHV) with minimal downtime and fully automated steps, reducing the operational burden.”

Increasing vCPUs would not address storage-level performance issues, and collapsing disks would require significant reconfiguration and VM downtime. vDisk sharding is not directly user-configurable and not applicable in this scenario.