Troubleshooting and Optimization

DVA-C02 · 54 questions

• Citizens report double charges on a fee API. Logs show the payment handler processing the same retried request twice without an idempotency check. What is the most likely code-level root cause to fix?
• Error rate for a city payments Lambda is flat, but p99 latency spiked. X-Ray shows a slow downstream payment provider call rather than elevated Lambda CPU. Where should the team focus first?
• After a citizen complaint, engineers must find all log events that mention a specific permitId across thousands of lines. Which approach extracts those events quickly?
• Business stakeholders want a CloudWatch metric for permits submitted, emitted from the application's structured logs, without flooding the API with separate PutMetricData calls. Which approach should developers implement?
• During a weather-emergency surge, a municipal ops desk opens one view that shows API Gateway 5xx rates, Lambda duration, and DynamoDB throttle counts together. What should they use to review that application health in one place?
• A city Elastic Beanstalk release through CodeDeploy fails. Service output logs show a failed health-check hook, not a compile error. What should the team conclude first?
• A municipal permit queue grows because an SQS-triggered Lambda’s batch size and visibility timeout do not match how long each batch takes to process. What is the most likely integration problem?
• Citizen profile PutItem calls fail while CloudWatch shows ProvisionedThroughputExceeded on the table. Developers suspected a null-pointer in app code. What does the evidence indicate?
• One citizen checkout spans API Gateway, Lambda, Step Functions, and DynamoDB, and support needs to see which hop failed. Which approach best finds the failing step?
• A library catalog Lambda is slow only on the first request after idle; later calls are fine. What latency pattern does that most strongly suggest?
• After an IAM role policy change, municipal Lambda logs show AccessDenied on s3:GetObject for a known object key. What is the most likely root cause?
• After a canary deployment, error budgets burn and logs/metrics implicate the new Lambda version. What observability-backed action should the team take?
• SigV4 signature failures cluster on one park-kiosk image whose clock is wrong via NTP, while other kiosks succeed. What environmental cause should developers check?
• Inspection events land in an SQS dead-letter queue. Opening the poison messages shows a newer JSON schema the consumer does not understand yet. What should analysis conclude?
• Right after a release, DynamoDB consumed write capacity roughly doubles at the same citizen traffic level. What should the team do with metrics?
• Support cannot stitch API Gateway access logs to Lambda log lines for one citizen complaint because nothing ties the records together. What should the app include?
• X-Ray traces mark some downstream calls as errors (client 4xx) and others as faults (5xx). Why does that classification matter for a municipal incident?
• A grants Lambda extracts ZIPs into /tmp; later invokes on the same warm environment fail in odd ways. What resource problem fits?
• Production suddenly looks empty: reads miss citizen rows that still exist. A pipeline parameter typo pointed the app at the test table name. What should troubleshooting check first?
• A partial outage plus aggressive client retries turns into DynamoDB throttling across the fleet. What contributing cause should developers address?
• A program manager says the permit platform already has logs, so it has full observability. How should engineers clarify the gap?
• For a building-permit workflow, which logging strategy best records behavior without drowning operators in noise?
• Default AWS metrics do not show how many permit jobs a worker cleared from a custom backlog. What should the code emit?
• During incidents, engineers need to filter traces by districtId and permitType. What should they add in the application?
• Ops wants SNS notices when account-level Lambda concurrency nears quota and when a production deploy finishes. What should they implement?
• A citizen request crosses API Gateway into Lambda, and the team wants a service map of that path. What should they enable?
• Operators struggle to query free-text Lambda logs for citizenId, action, and latencyMs. What logging change helps CloudWatch Logs Insights?
• An ALB target group for an ECS civic app should not receive traffic until DB connections are warmed and feature flags are loaded. What should they configure?
• Structured logs for a benefits app accidentally include Authorization headers and raw SSNs. What instrumentation rule was broken?
• A municipal payment portal must page on-call staff when HTTP 5xx rises and also when successful payment count stays at zero for fifteen minutes during office hours. How should the city developer design CloudWatch alerting?
• A high-volume city telemetry API cannot afford to record an X-Ray trace for every request without exploding cost. What sampling approach should the developer apply?
• A county Lambda in production writes DEBUG logs with infinite CloudWatch Logs retention, driving cost and widening exposure of sensitive permit fields. What should the developer change?
• Critical CloudWatch alarms for a city permitting API must reach the on-call municipal engineer quickly. Which wiring best matches how operators actually get paged?
• A citizen GIS lookup API is slow, but overall API duration mixes city code with a third-party GIS vendor. How should the developer instrument latency for clearer ownership?
• X-Ray service maps for a parks permit Lambda blur DynamoDB and outbound HTTP into the parent segment. What should the developer add for clearer remote call visibility?
• A municipal cache-warming process can keep a container process alive while it is still not ready to serve citizen traffic. How should health probes be designed?
• Operators need the same request identifier in API Gateway access logs, Lambda logs, and X-Ray traces for a single citizen complaint. What practice enables that correlation?
• CloudWatch custom metrics for a permitting API need actionable dashboards. Which dimension strategy avoids high-cardinality explosions?
• City audit chat requires a record whenever a production deploy finishes. How should the pipeline notify that completion?
• Citizen traffic risks being blocked when the city API approaches API Gateway throttle limits or STS session quotas. What should developers alarm on?
• During open enrollment, a benefits Lambda sees many overlapping invokes. How should the developer explain concurrency versus reserved concurrency?
• Profiling a receipt-generation Lambda shows JSON serialization dominating CPU on the hot path. What should the municipal developer do first?
• Load tests must find the smallest Lambda memory setting that still meets p95 latency for a city PDF-merge function. What is the right optimization goal?
• An SNS topic fans citywide alerts, but the parks department Lambda should only process messages tagged department=parks. How should invocations be reduced?
• A multi-lingual citizen portal on CloudFront must vary cache by Accept-Language without disabling caching entirely. What should the developer configure?
• Hot fee-schedule reads hammer Aurora during business hours for a city cashiering app. What application-level change reduces database load?
• CloudWatch shows consistent CPU under-utilization on an ECS OCR service for permit scans. What should the team do?
• p99 latency spikes on a warehouse Lambda align with many synchronous calls to a chatty inventory API. What remediation should the developer propose?
• Structured duration fields in application logs show an external geocoder call is about 80 percent of request time for address validation. What conclusion should operators draw?
• A meter-event consumer invokes Lambda once per SQS message, wasting per-invoke overhead. How should processing be optimized?
• A permit handler runs five separate DynamoDB queries when one keyed access pattern with a wider item (or single query) would suffice. What should the developer change?
• A city Lambda pays TCP/TLS handshake costs on every invoke when creating new HTTP and AWS SDK clients inside the handler. What reuse pattern helps?
• Citizen photo uploads feel slow because thumbnail generation runs on the synchronous upload API path. How should latency be improved?
• An open-data catalog can tolerate roughly two minutes of cache staleness, but election-night results pages need near-zero stale cache. How should cache TTLs be set?