Troubleshooting practice questions

Horizontal Pod Autoscaler configuration

Why wrong: HPA controls replica count scaling, not DNS resolution.

Persistent Volume Claim binding state

Why wrong: PVC binding relates to storage, not DNS name resolution.

Ingress controller SSL certificate expiration

Why wrong: SSL certificates affect TLS handshakes, not raw DNS lookups.

Hardcode the admin username and password into environment variables

Why wrong: Hardcoding credentials is a security anti-pattern and violates security best practices.

Expose the ACR endpoint publicly without authentication

Why wrong: ACR requires authentication and does not support unauthenticated public pull without specific anonymous artifact configurations.

Attach an Azure Storage Account SAS token to the container environment

Why wrong: SAS tokens are for Azure Storage, not container registry image pulls.

Configuring Vertical Pod Autoscaler.

Why wrong: VPA manages resource requests, not container startup speed.

Increasing the memory limit for the pod.

Why wrong: Increasing memory limits does not speed up the container pull process.

Using a smaller base OS distribution.

Why wrong: While this helps, Artifact Streaming is a specific feature for startup optimization.

Kubernetes standard kube-proxy in IPVS mode only

Why wrong: kube-proxy handles service routing and load balancing, not pod-to-pod network security policy enforcement.

Azure Virtual Network NAT Gateway

Why wrong: NAT Gateway manages outbound SNAT connectivity to the public internet, not internal pod communication.

Kubenet basic networking without network policies

Why wrong: Kubenet does not support native Azure network policies.

Configuring high-frequency indexers

Why wrong: Indexer frequency affects CPU and IOPS during ingestion, not steady-state index memory consumption.

Enabling semantic ranking on text fields

Why wrong: Semantic ranking consumes CPU and memory for the inference models, but the primary RAM consumer for vector search is the HNSW graph index itself.

Using exhaustiveKnn instead of HNSW

Why wrong: Exhaustive KNN typically uses less memory than HNSW because it does not maintain multi-layer graph structures in RAM, though search is slower.

Convert all vector fields to Edm.String

Why wrong: Converting vector fields to strings breaks vector indexing entirely.

Switch the search service to the Free tier

Why wrong: The Free tier has strict time limits and would cause indexers to fail faster.

Delete all semantic configurations from the index

Why wrong: Semantic configurations affect query time ranking, not indexer document parsing timeouts.

Replace all vector fields with standard ASCII folding tokenizers

Why wrong: ASCII folding tokenizers are for text normalization, which would break vector search entirely.

Configure the index to use exhaustive KNN exclusively for high-QPS production workloads

Why wrong: Exhaustive KNN reduces query throughput at scale; HNSW should be used for high-QPS production workloads.

Change the Blob storage tier to Cool.

Why wrong: Storage tier does not impact the compute-intensive nature of the skillset execution phase.

Switch from HNSW to Flat indexing.

Why wrong: Switching to Flat will likely increase search latency and is not related to ingestion bottlenecks.

Reduce the embedding vector dimensions.

Why wrong: Changing dimensions is not a configuration setting; it requires re-generating embeddings for all data.

Deploy an Azure VPN Gateway on the local workstation.

Why wrong: An enterprise VPN gateway is an over-engineered infrastructure solution when a simple client configuration change (WebSockets) solves the port issue.

Switch the connection string to use an Event Grid webhook protocol.

Why wrong: Event Hubs producers/consumers require AMQP or Kafka protocols, not Event Grid webhooks.

Open an Azure Support ticket to whitelist port 5671 on the Azure backbone router.

Why wrong: Azure backbone routing is not the bottleneck; the corporate firewall is blocking outbound 5671.

Enabling Service Bus message session locks on the event subscription

Why wrong: Service Bus sessions do not apply to Event Grid subscriptions.

Running the az servicebus queue purge command

Why wrong: Service Bus CLI commands do not manage Event Grid subscriptions.

Using Event Hubs Capture to replay webhook HTTP headers

Why wrong: Capture archives stream data, it does not manage Event Grid webhooks.

Verify the AMQP listener port 5671 firewall rules on the function app.

Why wrong: Event Grid webhooks communicate over HTTP/HTTPS, not AMQP.

Review Service Bus session lock renewal logs.

Why wrong: The scenario involves Event Grid and Azure Functions, not Service Bus sessions.

Check Event Hubs consumer group checkpoint offset status.

Why wrong: Event Hubs checkpoints do not apply to Event Grid triggers.

Azure Cost Management amortization tables

Why wrong: Incorrect. Amortization tables track financial cost distribution, not authentication errors.

Azure Storage blob access audit logs

Why wrong: Incorrect. Storage logs only track blob operations, not Azure AD token endpoint requests.

Azure Key Vault secret version history

Why wrong: Incorrect. Key Vault tracks secret versions, not Azure AD token generation failures.

Azure Network Watcher.

Why wrong: Network Watcher is for connectivity and flow analysis, not disk performance.

Azure Resource Graph Explorer.

Why wrong: Resource Graph is for querying resource metadata, not performance metrics.

Application Insights Dependency logs.

Why wrong: Dependency logs track service-to-service calls, not infrastructure-level disk IO.

Kubenet with host-port mapping

Why wrong: Host-port mapping binds container ports directly to node ports, limiting scalability and routing flexibility.

Kubenet with basic Linux bridge

Why wrong: Kubenet uses NAT and route tables, introducing additional routing complexity and overhead.

Azure CNI Overlay

Why wrong: Overlay networking encapsulates traffic using Geneve, adding overhead compared to direct VNet IP allocation.

In the Azure Container Registry webhook configuration

Why wrong: ACR webhooks trigger events on repository actions, unrelated to runtime resource limits.

In the Azure Virtual Network route table settings

Why wrong: Route tables control subnet traffic routing, not container compute limits.

In the Azure Monitor diagnostic settings export destination

Why wrong: Diagnostic settings configure log routing, not compute sizing.

Azure Cosmos DB request unit (RU) consumption logs

Why wrong: Incorrect. Cosmos DB logs track database throughput, not container console execution errors.

Azure Key Vault audit logs

Why wrong: Incorrect. Key Vault audit logs track secret access, not application container console output.

Azure Advisor performance recommendations

Why wrong: Incorrect. Advisor gives high-level guidance, not real-time container crash logs.

Azure Bastion connection to the container host VM

Why wrong: Azure Container Apps abstracts underlying hosts entirely; direct VM access via Bastion is not available.

AKS control plane audit logs

Why wrong: AKS audit logs apply to Kubernetes clusters, not Azure Container Apps.

Azure Storage Account blob container metrics

Why wrong: Blob storage metrics do not contain container application console logs or health probe results.

Execute az aks browse to open the Kubernetes dashboard.

Why wrong: This command applies to AKS clusters, not Azure Container Apps.

Use kubectl describe pod to inspect Kubernetes events.

Why wrong: Azure Container Apps abstracts away the underlying Kubernetes control plane, so direct kubectl commands against container app pods are not supported.

Modify Azure subscription billing policies to cap Log Analytics data ingestion at zero cost

Why wrong: Incorrect. Capping workspace ingestion drops data arbitrarily or stops logging entirely rather than managing sampling intelligently.

Delete the Log Analytics workspace and recreate it with a lower pricing tier

Why wrong: Incorrect. Deleting the workspace deletes all stored historical log data and does not dynamically adjust application sampling rates.

Use Azure Policy to deny all incoming Application Insights trace requests

Why wrong: Incorrect. Azure Policy does not control application-level telemetry sampling rates.

Virtual machine CPU core count assigned to the client application

Why wrong: Client VM CPU does not affect server-side search relevance algorithms.

Azure Blob Storage container access tier (Hot vs Cool)

Why wrong: Storage access tiers affect blob storage costs, not search relevance scoring.