Practice for the Red Hat Certified Specialist in OpenShift AI (EX267) exam with 18 exam-style practice questions, instant answer reveals, and concise explanations of every correct answer. Topics include: Which method grants secured Git repository authentication systems to re-use protected code inside workbenches seamlessly. Follow @CertPunch and visit certpunch.com for more certification practice exams and study content.
Prefer hands-on? Take this round as an interactive practice test — answer every question, get instant feedback, and see your score: Start the Red Hat Certified Specialist in OpenShift AI (EX267) practice test →
What you will practice
- Which method grants secured Git repository authentication systems to re-use protected code inside workbenches…
- Which operator manages the cluster ingress traffic capabilities that OpenShift Service Mesh uses to find mode…
- What occurs if a pipeline step attempts to allocate more CPU resources than permitted by the namespace Resour…
- What component converts an Elyra pipeline topology map visual layout into a clean Tekton-compliant YAML manif…
- WHAT CRD SETS THE GLOBAL DEFAULTS FOR USER CONFIGURATIONS LIKE NOTEBOOK CULLING TIMEOUTS VALUE?
- Which format is utilized to write custom metric monitors for ingestion by the TrustyAI analysis engine servic…
Answers and explanations
Tap a question to expand the answer and the exam reasoning. Try to commit to your own pick first.
Q1. Which method grants secured Git repository authentication systems to re-use protected code inside workbenches seamlessly?
Answer: D. Injecting SSH Private Keys or Personal Access Tokens via Kubernetes Secrets
Workbenches consume Git credentials seamlessly by mounting SSH keys or personal access tokens via Kubernetes Secrets. Plaintext passwords in configmaps are highly insecure and are not natively supported by the workbench interface.
Q2. Which operator manages the cluster ingress traffic capabilities that OpenShift Service Mesh uses to find model endpoints?
Answer: C. OpenShift Ingress Operator or Service Mesh Ingress Gateway controllers
The OpenShift Ingress Operator and Service Mesh Ingress Gateway handle incoming cluster traffic and route it appropriately. GitOps or Data Foundation operators manage continuous delivery and storage, bypassing ingress routing duties.
Q3. What occurs if a pipeline step attempts to allocate more CPU resources than permitted by the namespace ResourceQuota?
Answer: C. The pipeline task pod remains in a Pending state and eventually times out
If a pod requests more CPU than the namespace ResourceQuota allows, Kubernetes leaves it in a Pending state. OpenShift AI relies on these native constraints, so tasks will not execute on master nodes or bypass limits when quotas are exceeded.
Q4. What component converts an Elyra pipeline topology map visual layout into a clean Tekton-compliant YAML manifest definition?
Answer: A. Elyra Kubeflow/Tekton runtime processor plugin modules
The Elyra runtime processor plugin translates the visual pipeline topology into the Tekton-compliant YAML manifest. The OpenShift Router and storage drivers handle network and volume operations, entirely unrelated to pipeline code compilation.
Q5. WHAT CRD SETS THE GLOBAL DEFAULTS FOR USER CONFIGURATIONS LIKE NOTEBOOK CULLING TIMEOUTS VALUE?
Answer: D. DashboardConfig
The DashboardConfig custom resource defines user-facing settings like notebook culling timeouts and other dashboard defaults. DataScienceCluster manages installed components, but it does not control user session or culling configurations.
Q6. Which format is utilized to write custom metric monitors for ingestion by the TrustyAI analysis engine service?
Answer: C. Structured JSON or YAML definition manifests defining the metric target parameters
Custom metric definitions for TrustyAI are defined using structured JSON or YAML manifests to specify the required metric parameters. The service consumes declarative configurations rather than reading unstructured text or executing binary files.
Q7. What status category in the InferenceService CR confirms a model is fully ready to answer inference traffic?
Answer: D. Ready condition set to True within the resource status block
When you check an InferenceService resource, the Ready condition must be True to confirm it accepts traffic. Ignore deployment logs; always rely on the specific Kubernetes status block for readiness.
Q8. What happens to a deployed model serving instance if the cluster experiences a total loss of connection to the underlying S3 bucket?
Answer: D. The model remains serving existing traffic from memory but cannot reload or update
The runtime keeps the model in memory, so existing predictions continue working without interruption. However, it fails to fetch updates or reload if the pod restarts, since storage connectivity is required.
Q9. What happens if you delete a pipeline definition template from the RHOAI dashboard while a pipeline run is executing?
Answer: D. The active pipeline run continues executing to completion using its cached execution manifest copy
Active pipeline executions rely on an independent runtime manifest, so deleting the source template does not interrupt them. The Argo workflow engine tracks the live execution separately from the dashboard template.
Q10. Where can you configure global proxy settings to ensure RHOAI components can reach external internet model repositories?
Answer: D. Inside the DSCInitialization custom resource configuration specifications
Global proxy settings for cluster-wide RHOAI component egress are configured inside the DSCInitialization custom resource. Configuring individual user notebook menus only applies proxy variables locally and fails to provide platform-wide external repository access.
Q11. Which graphical dashboard display interface helps enterprise workers track TrustyAI bias analysis evaluations visually?
Answer: B. Model Serving analytics screens or dedicated TrustyAI dashboards metrics within OpenShift AI UI
TrustyAI bias metrics are visualized directly through the Model Serving analytics screens or dedicated TrustyAI dashboards within the OpenShift AI interface. External telemetry tools like Grafana monitor general cluster health and lack built-in, specialized bias evaluation views.
Q12. What environment variable holds the target service destination path when configuring a pipeline step to send data to an external API?
Answer: D. Custom developer-defined environment variable strings inside step properties
When sending pipeline data to an external API, developers must define custom environment variables within the specific step properties to hold the destination path. System-reserved variables like ODS_DATA_CONNECTION are strictly managed by the platform for internal storage connections.
Q13. What method reduces total deployment overhead costs for teams managing multiple compact machine learning infrastructures outside LLMs?
Answer: C. Deploying models into Multi-Model Serving configurations utilizing ModelMesh tools
Deploying multiple compact models into Multi-Model Serving configurations utilizing ModelMesh significantly reduces infrastructure overhead by sharing pod resources. Provisioning separate single-model inference systems wastes compute resources and increases operational management costs for smaller models.
Q14. What configuration specifies the maximum time a Knative serverless model container can remain idle before scaling to zero?
Answer: A. scale-to-zero-grace-period configuration value
The scale-to-zero-grace-period setting dictates the waiting time before Knative terminates an idle serverless inference pod. Rely on standard Knative terminology here, ignoring distractors that sound plausible but are not actual configuration parameters.
Q15. What command downloads diagnostic event details specific to Red Hat OpenShift AI core controller deployment crashes?
Answer: B. oc get events -n redhat-ods-applications –sort-by='.metadata.creationTimestamp'
Querying OpenShift events in the redhat-ods-applications namespace reveals deployment issues and crash loops for the core controller. Remember to sort by timestamp to trace events chronologically during your troubleshooting process.
Q16. Which protocol routing layer controls scale-to-zero modes and delivers fast incoming request traffic for single-model systems?
Answer: C. Knative Serving platform orchestration routing modules and components architectures
Knative Serving manages network routing and automatically scales inactive inference pods down to zero to conserve resources. Standard OpenShift Routers lack this native scale-to-zero awareness, making them incorrect for serverless traffic management.
Q17. What type of model format is typically generated by frameworks like PyTorch and can be directly served via the Triton Inference Server?
Answer: C. TensorRT Plan, TorchScript, or ONNX format configurations
Triton Inference Server natively supports optimized formats like TorchScript, ONNX, and TensorRT plans. Pickle files are generally insecure and not directly supported for production model serving.
Q18. What type of persistent volume provisioning is typically recommended for storing large machine learning models locally on the model server?
Answer: B. High-performance block storage volumes (ReadWriteOnce)
Block storage configured as ReadWriteOnce provides the high IOPS required for local model storage. Ephemeral storage options like emptyDir risk data loss during pod restarts, making them unsuitable for persistent model files.
More Red Hat Certified Specialist in OpenShift AI (EX267) drills and other practice exams are on @CertPunch. New rounds drop every few days at certpunch.com.