NCP-AIO Installation and Deployment Practice Question
An administrator is deploying NVIDIA AI Enterprise on a Kubernetes cluster and must decide how GPU workloads should request accelerators. The environment has a mix of full-GPU training jobs and inference services that share a single A100. Which approach correctly allows a pod to consume a specific MIG-backed slice rather than the whole device?
⚠ Common exam trap
The trap here is believing that annotations or environment variables can steer GPU allocation, when only extended resource requests in the pod spec determine what the device plugin hands out.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Request the specific extended resource, such as nvidia.com/mig-2g.10gb, in the pod's resource limits.
In Kubernetes, GPU allocation is expressed exclusively through extended resource requests handled by the NVIDIA device plugin. When MIG is enabled, each configured profile is advertised under a distinct resource name, so a pod requesting a specific MIG profile is bound to a matching instance. This is what enables a single A100 to serve both whole-device training and sliced inference workloads.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Request nvidia.com/gpu: 1 and add the annotation nvidia.com/mig-profile=2g.10gb to the pod metadata.
Why it's wrong here
A request for nvidia.com/gpu allocates an entire physical GPU, and Kubernetes annotations are not consumed by the device plugin for allocation decisions. The annotation would be ignored, so the pod would receive a whole A100, defeating the goal of sharing the device between training and inference workloads.
- ✗
Set the environment variable NVIDIA_MIG_PROFILE=2g.10gb in the container spec and request nvidia.com/gpu: 1.
Why it's wrong here
The device plugin does not read container environment variables when deciding which resource to allocate. Environment variables can influence runtime behavior after allocation, but scheduling is driven solely by extended resource requests, so this combination still yields a full physical GPU rather than a MIG slice.
- ✗
Deploy a separate RuntimeClass named mig-2g.10gb and reference it from the pod's spec.runtimeClassName.
Why it's wrong here
RuntimeClass selects which container runtime handler processes the pod, such as the NVIDIA runtime for GPU injection. It does not describe GPU partitioning geometry and carries no MIG profile semantics, so referencing it cannot cause the scheduler or device plugin to hand the pod a specific MIG instance.
- ✓
Request the specific extended resource, such as nvidia.com/mig-2g.10gb, in the pod's resource limits.
Why this is correct
The NVIDIA device plugin advertises each configured MIG profile as its own extended resource name. A pod that requests nvidia.com/mig-2g.10gb in its limits is scheduled onto a node with a free instance of exactly that profile, giving the inference service a dedicated slice while full-GPU training jobs use other devices.
About these practice questions
Courseiva writes every NCP-AIO question from scratch — 309 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official NVIDIA exam blueprint
This NCP-AIO practice question is part of Courseiva's free NVIDIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the NCP-AIO exam.