Most "DVA-C02 exam questions and answers" posts are dumps in a trench coat: a wall of questions, single-letter answers, zero reasoning. That teaches you nothing, because the real AWS Certified Developer – Associate exam rewrites every scenario. So here are five worked examples instead — realistic question shapes with the reasoning spelled out — because each one encodes a pattern that covers a dozen exam questions.
If you want to drill this style at volume, the free DVA-C02 practice test is the no-signup place to start; every question there ships with the same kind of explanation.
Q1 — Lambda + SQS: the poison message
A Lambda function processes messages from an SQS queue. One malformed message causes the function to error repeatedly, blocking the batch. What's the best fix?
A) Increase the visibility timeout B) Configure a dead-letter queue with maxReceiveCount C) Increase Lambda memory D) Switch to long polling
Answer: B. The pattern being tested is failure isolation. A DLQ with maxReceiveCount shunts the poison message aside after N attempts so the rest of the queue flows. Visibility timeout (A) just changes retry timing; C and D are irrelevant. The transferable rule: whenever a question says "one bad record blocks processing," look for DLQ (SQS/Lambda destinations) or bisect-on-error (Kinesis).
Q2 — DynamoDB: hot partition on a popular key
A gaming leaderboard writes all scores for today's tournament to partition key TOURNAMENT#2026. Writes are throttled. What do you do?
A) Switch to on-demand capacity B) Add a GSI C) Shard the partition key with a random suffix D) Enable DAX
Answer: C. On-demand (A) is the trap — it raises the table's ceiling but a single partition key still caps out (~1,000 WCU/s per partition). Write sharding (TOURNAMENT#2026#7) spreads load; you fan-in on read. DAX (D) helps reads, not writes. Rule: throttling on one key = key design problem, not capacity problem. DVA-C02 asks this in at least three costumes.
Q3 — API Gateway + Lambda: cold-start latency
A REST API has p99 latency spikes on infrequently used endpoints. The team wants predictable latency without over-provisioning. Best option?
A) Provisioned concurrency with application auto scaling B) Bigger memory C) Keep-warm CloudWatch pings D) Move to EC2
Answer: A. Provisioned concurrency is the cold-start answer, and pairing it with auto scaling addresses "without over-provisioning." Keep-warm pings (C) is the legacy hack the exam wants you to reject. Memory (B) shortens cold starts but doesn't eliminate them. Rule: "predictable latency" + Lambda → provisioned concurrency; "cheaper cold starts" for Java → SnapStart.
Q4 — Security: credentials in code
A developer finds a database password committed in an environment variable in a Lambda function's configuration. Compliance requires rotation. Where should it live?
A) SSM Parameter Store (SecureString) B) Secrets Manager C) KMS directly D) Encrypted S3 object
Answer: B. The keyword is rotation — Secrets Manager does managed rotation natively; Parameter Store SecureString (A) is the right answer only when the question emphasizes free/simple and omits rotation. KMS (C) encrypts data, it doesn't store secrets. Rule: rotation → Secrets Manager; cheap config → Parameter Store. This single distinction is worth multiple points on most forms.
Q5 — Deployment: safe Lambda release
A team wants to ship a new Lambda version to 10% of production traffic and auto-rollback on errors. Which combination works?
A) Lambda alias + CodeDeploy canary deployment with CloudWatch alarms B) Blue/green with two API Gateways C) Weighted Route 53 records D) Deploy and watch logs
Answer: A. Aliases give you traffic shifting between versions; CodeDeploy's Canary10Percent configs plus alarm-triggered rollback is exactly the described requirement. Rule: the DVA-C02 loves the alias/version model — know that $LATEST is unpublished, aliases point at versions, and traffic shifting happens at the alias.
What these five have in common
Notice what the exam is actually testing: not trivia, but decision rules — rotation vs cheap, key design vs capacity, isolation vs retry. The current DVA-C02 blueprint spreads this across four domains (development ~32%, security ~26%, deployment ~24%, troubleshooting/optimization ~18%), and security is the domain that most often sinks developers who build daily but never touch IAM policy conditions or KMS encryption context.
A workflow that reliably gets people from "I use AWS at work" to passing in 4–6 weeks:
- Baseline with 25 fresh questions; note your per-domain scores.
- For every miss, write the one-line rule (like the bold ones above) into a running list.
- Drill 25/day from a bank that explains answers — the ExamCert DVA-C02 page is built for exactly this loop, and when an explanation still doesn't land you can push the same question into the AI simulator at https://ai.examcert.app and ask follow-ups until it does.
- Two timed full-lengths in your final week; book the exam when fresh-question accuracy clears 80%.
Sixty-five questions, 130 minutes, ~720/1000 to pass. The candidates who fail are rarely short on AWS experience — they're short on rules. Collect fifty of them and the exam becomes pattern-matching you've already done.
Top comments (0)