A data science team reports that their Jupyter notebook pod, which was previously working, is now failing to start. The pod's status is 'CrashLoopBackOff'. An MLOps engineer investigates and finds that the pod's PersistentVolumeClaim (PVC) is bound, but the pod logs show a "Permission denied" error when trying to write to its '/data' mount point.
The engineer checks the Trident backend configuration associated with the pod's StorageClass:
apiVersion: trident.netapp.io/v1
kind: TridentBackendConfig
metadata:
name: ontap-nas-eco
spec:
version: 1
storageDriverName: ontap-nas
managementLIF: 10.10.20.5
dataLIF: 10.10.20.10
svm: svm-prod-ds
exportPolicy: read-only-policy
What is the most likely cause of the "Permission denied" error?

• A. The 'storageDriverName' should be 'ontap-san' for all AI workloads.
• B. The Trident backend is configured to use an export policy ('read-only-policy') that does not grant write permissions to the Kubernetes nodes.
• C. The 'dataLIF' is configured incorrectly and is unreachable from the Kubernetes nodes.
• D. The Kubernetes pod has an invalid 'securityContext' that prevents it from writing to any volume.

The firm's data science team needs to run a high-priority, interactive model analysis job that requires immediate access to two GPUs. However, all GPUs in the cluster are currently allocated to long-running, lower-priority batch training jobs.
The MLOps platform, Run:AI, shows the following queue status:
JOB_ID | PROJECT | STATUS | PRIORITY | GPU_ALLOCATED
||--|-|
batch_job_1 | team_a | Running | Low | 2
batch_job_2 | team_a | Running | Low | 2
batch_job_3 | team_b | Running | Low | 4
interactive_1| team_c | Pending | High | 2 (requested)
How does the Run:AI platform address this resource contention to allow the high-priority job to run?

• A. It automatically pauses one of the low-priority jobs, saves its state, and allocates its GPUs to the high- priority job, placing the paused job back in the queue.
• B. It sends an email notification to the administrator to manually reallocate the GPUs.
• C. It keeps the high-priority job in a pending state until the low-priority jobs complete naturally.
• D. It automatically terminates all low-priority jobs to free up the entire cluster.

The firm wants to extend the "Advisor Assistant" to include a new batch processing feature. Every night, the system must analyze every client portfolio against a set of 50 different risk models and generate a compliance report. This is a highly parallel, read-intensive workload. The architect must design a data workflow that is efficient and does not impact the production chatbot environment. Which sequence of actions and technologies provides the most effective solution?

• A. Create a full physical copy of the client portfolio database to a separate volume, mount it to the compute nodes, and run the analysis.
• B. Create a NetApp Snapshot of the portfolio database volume, create a FlexClone from that snapshot, mount the FlexClone to the analysis pods, and run the batch job.
• C. Use NetApp SnapMirror to replicate the portfolio database volume to the DR site, and run the analysis jobs there.
• D. Run the analysis job directly against the production portfolio database during off-peak hours.

An MLOps engineer is deploying a training pod that requires a high-performance volume. After applying the pod and PVC manifests, the pod remains in a 'Pending' state. The engineer runs
'kubectl describe pod training-pod-7d8c' and sees the following event:
Events:
Type Reason Age From Message
- - - -
Warning FailedScheduling 2m15s
default-scheduler 0/4 nodes are available: 1 node(s) had volume node affinity conflict, 3 node(s) didn't find available persistent volume to bind.
The engineer then inspects the associated PVC and sees its status is also 'Pending'.
What is the most likely cause of this issue?

• A. The 'storageClassName' specified in the PVC does not match any existing StorageClass managed by Trident.
• B. The ONTAP backend is out of available capacity to provision a new volume.
• C. The Kubernetes scheduler is malfunctioning and cannot assign pods to nodes.
• D. The Trident controller pod has crashed and needs to be restarted.