DVA-C02 AWS Certified Developer - Associate Questions and Answers

A function alias is a pointer to a specific Lambda function version. You can use aliases to create different environments for your function, such as development, testing, and production. You can also use aliases to perform blue/green deployments by shifting traffic between two versions of your function gradually. This way, you can easily roll back to a previous version if something goes wrong, without having to redeploy your code or change your configuration. Reference: AWS Lambda function aliases

Why Option C is Correct:Implementing a connection pool in the Redis client reduces connection overhead and avoids frequent connection establishment, which helps to reduce latency and connection failures.

Why Other Options are Incorrect:

Option A: Exponential backoff retry strategies help with transient failures but do not resolve latency caused by frequent connection establishment.

Option B: Storing scores in application memory adds complexity and risks inconsistency.

Option D: Adding more nodes is unnecessary unless the cluster is under heavy load. Latency due to connection failures is better addressed at the application level.

AWS Documentation References:

Amazon ElastiCache Best Practices

The correct answer is B because Amazon S3 event notifications can directly invoke an AWS Lambda function whenever a supported event occurs, such as ObjectCreated . This is the simplest and most infrastructure-efficient design for processing files that are uploaded to an S3 bucket. Since the requirement is to invoke the function with the least amount of AWS infrastructure , using the native direct integration between S3 and Lambda is the best choice.

AWS documentation explains that S3 can publish event notifications for object-level operations and send those notifications directly to Lambda. This allows the developer to build an event-driven workflow in which each newly uploaded project status file automatically triggers processing logic. There is no need for polling, scheduled scans, or intermediary messaging services unless there are additional requirements such as buffering, fan-out, or decoupling.

Option A is incorrect because polling the bucket every 5 minutes with EventBridge is less efficient and adds unnecessary delay and complexity. Option C is incorrect because introducing an SNS topic adds another infrastructure component when direct invocation is already supported. Option D is also incorrect because using SQS introduces a queue and polling behavior, which increases operational overhead and infrastructure count.

The direct S3-to-Lambda pattern is a common AWS serverless best practice for file processing workloads. It minimizes moving parts, reduces latency, and simplifies permissions and maintenance. Therefore, the developer should configure an S3 event notification for object creation events to invoke the Lambda function directly.

So, B is the correct answer.

A common performance issue for Lambda functions that access relational databases is the overhead of establishing a new database connection on every invocation . In the provided pseudocode, each call to lambda_handler opens a connection, executes one statement, and closes the connection. When the function runs hundreds of times per hour , repeatedly creating and tearing down connections adds latency (TCP handshake, TLS negotiation if used, authentication, database session setup), and can also increase load on the database connection manager.

AWS guidance for Lambda execution environments explains that the runtime environment may be reused across multiple invocations of the same function (a “warm” execution environment). Code that is initialized outside the handler runs once per execution environment lifecycle and can be reused on subsequent invocations in that environment. By moving database.connect() into the global scope (module-level initialization), the function can reuse an existing connection when the execution environment is warm, reducing per-invocation latency and improving overall throughput.

This change directly targets the slow part of the flow (connection setup) without changing the database engine, resizing the database, or adding unrelated concurrency settings. Increasing reserved concurrency (A) controls scaling limits but does not reduce the per-invocation connection overhead; it can even amplify connection storms. Resizing the database (B) might help capacity, but it does not address the repeated connection establishment cost and is not the first optimization. Replacing RDS with DynamoDB (D) is a major redesign and unnecessary for the stated issue.

So the best improvement, with minimal code change and aligned with Lambda best practices, is option C : create the connection outside the handler and reuse it when possible, while still handling retries and reconnect logic if the connection becomes stale.

The requirements are: (1) store an API key that a Lambda function will use, (2) ensure the secret is encrypted at rest using AWS KMS , and (3) the company must control key rotation , which implies using a customer managed KMS key (CMK) rather than an AWS managed key.

Option C meets the requirements by storing the API key in AWS Systems Manager Parameter Store as a SecureString parameter encrypted with a customer managed KMS key . SecureString is designed for sensitive configuration data and integrates with KMS so the organization can choose the CMK, manage its lifecycle, and control rotation policies. The Lambda function can retrieve the parameter at runtime using the AWS SDK, and IAM policies can tightly control access to the parameter and to the KMS key.

Option E also meets the requirements by storing the API key in a Lambda environment variable encrypted with a customer managed KMS key . Lambda encrypts environment variables at rest and allows you to specify a customer managed KMS key for encryption. This gives the company control over key rotation and key policy, satisfying the “must control key rotation” requirement. The function reads the value from the environment at runtime without additional network calls.

Why the other options fail:

A uses an AWS managed KMS key, which does not satisfy the requirement for the company to control rotation (you cannot manage rotation of AWS managed keys in the same way).

B is a plain String parameter, which is not encrypted as a secret and does not meet the at-rest encryption requirement.

D uses an AWS managed KMS key, again failing the company-controlled rotation requirement.

Therefore, the two valid solutions are C (Parameter Store SecureString with a customer managed CMK) and E (Lambda environment variable encrypted with a customer managed CMK).

AWS X-Ray SDK: The primary way to enable X-Ray tracing within applications. The SDK sends data about requests and subsegments to the X-Ray daemon for service map generation.

Instrumenting All Services: To visualize a complete microservice architecture on the service map, each relevant service must include the X-Ray SDK.

The correct answer is D because the current DynamoDB table design is causing a hot partition problem. The table uses order_date as the partition key, which means that during peak ordering periods, a very large number of writes are likely targeting the same partition key value for the current date. In DynamoDB, write throttling often occurs when traffic is not evenly distributed across partition keys, even when the table uses on-demand mode . On-demand capacity can automatically scale, but it still depends on proper key distribution to avoid concentrated activity on a small number of partitions.

By changing the partition key to customer_id and using order_id as the sort key, writes will be spread across many customers instead of being concentrated on a small number of date values. This creates a more balanced write pattern and better aligns with DynamoDB best practices for high-scale workloads. DynamoDB documentation emphasizes choosing partition keys with high cardinality and even request distribution to avoid throttling and improve performance.

Option A is not a good design because a sequential partition key can still create uneven access patterns and usually does not support meaningful access requirements. Option B is not the best answer because simply increasing capacity does not fix a poor partition key design; hot partitions can still throttle. Option C is unnecessary because the issue is not that DynamoDB is the wrong service, but that the table schema needs redesign.

Therefore, the best solution is to refactor the key schema so that writes distribute more evenly across partitions, which is exactly what D achieves.

S3Bucket and S3Key: These properties in a CloudFormation AWS::Lambda::Function resource specify the location of the function ' s code in S3.

Least Development Effort: This solution minimizes code changes, relying on CloudFormation to reference the existing S3 deployment package.

For client-side encryption with AWS KMS, the documented pattern is envelope encryption : you use KMS to generate and protect a data encryption key (DEK) , and you use that DEK locally to encrypt the actual data. This is the correct approach for large payloads (such as 10 MB documents), because the KMS Encrypt API is intended for encrypting small amounts of data (KMS is not designed to directly encrypt multi-megabyte application payloads).

With envelope encryption, the application calls GenerateDataKey on AWS KMS, specifying the KMS key (customer managed key) to use. KMS returns two versions of the DEK in the response:

a plaintext DEK (usable immediately by the application for local cryptographic operations), and

a ciphertext (encrypted) DEK that is encrypted under the specified KMS key.

The application then uses the plaintext DEK to encrypt the 10 MB document on the client side (for example, using an authenticated encryption mode such as AES-GCM via a standard crypto library). After encryption completes, the application must not persist the plaintext DEK; instead, it stores the encrypted document alongside the ciphertext DEK . Later, to decrypt, the application sends the ciphertext DEK to KMS using Decrypt to recover the plaintext DEK, and then decrypts the document locally.

Therefore, option D is correct because it explicitly uses GenerateDataKey and the plaintext DEK to encrypt the data, which is the required envelope-encryption flow. Options A, B, and C do not follow the correct KMS envelope-encryption process for large client-side payloads.

Requirement Summary:

Lambda function accesses RDS for MySQL

Credentials are currently hardcoded in code (insecure)

Must enable automated credential rotation every 30 days

Option A: Use AWS Secrets Manager + automatic rotation

Best and secure option

Secrets Manager allows:

Secure storage of secrets

Integration with RDS for automatic rotation

Scheduled rotation every X days (e.g., 30 days)

Lambda can fetch credentials via SDK (GetSecretValue)

Option B: Use SSM Parameter Store + rotation

SSM does not support automatic rotation of secrets.

You ' d need to build custom rotation logic = higher operational overhead.

Option C: Encrypted S3 + Object Lambda rotation

Not intended for credential storage.

S3 is not a secrets management system, and Object Lambda does not perform rotation.

Option D: SSM + EventBridge rotation

No native integration between Parameter Store and EventBridge for secret rotation.

You’d need to build custom Lambda functions = higher maintenance.

Secrets Manager rotation for RDS: https://docs.aws.amazon.com/secretsmanager/latest/userguide/rotating-secrets.html

Secure retrieval in Lambda: https://docs.aws.amazon.com/secretsmanager/latest/userguide/retrieving-secrets_lambda.html

Integration with RDS MySQL: https://docs.aws.amazon.com/secretsmanager/latest/userguide/integrating_rds_config.html

This solution will meet the requirements by using IAM database authentication for Aurora, which enables using IAM roles or users to authenticate with Aurora databases instead of using passwords or other secrets. The developer can use IAM database authentication for Aurora to enable secure database connections for all the Lambda functions that access Aurora DB instance. The developer can create an IAM role with permission to connect to Aurora DB instance and attach it to each Lambda function. The developer can also configure Aurora DB instance to use IAM database authentication and enable encryption in transit using SSL certificates. This way, the Lambda functions can use a single securely encrypted database connection string to access Aurora without needing any secrets or passwords. Option B is not optimal because it will store the credentials and read them from an encrypted Amazon RDS DB instance, which may introduce additional costs and complexity for managing and accessing another RDS DB instance. Option C is not optimal because it will store the credentials in AWS Systems Manager Parameter Store as a secure string parameter, which may require additional steps or permissions to retrieve and decrypt the credentials from Parameter Store. Option D is not optimal because it will use Lambda environment variables with a shared AWS Key Management Service (AWS KMS) key for encryption, which may not be secure or scalable as environment variables are stored as plain text unless encrypted with AWS KMS.

Comprehensive and Detailed Step-by-Step Explanation:

Option A: Use Amazon S3 for Shared Test Events:

Storing JSON test event files in an S3 bucket provides a centralized, cost-effective, and highly available solution.

Granular IAM policies can restrict access to specific developers or roles, ensuring security while maintaining consistency for shared test events.

This solution has minimal operational overhead and integrates easily with existing workflows.

Why Other Options Are Incorrect:

Option B: Using DynamoDB and a Lambda function introduces unnecessary complexity for a relatively simple requirement. S3 provides a simpler and more cost-efficient solution.

Option C: AWS Lambda test events are not inherently shareable across developers, making this option invalid.

Option D: Using a Git repository adds operational overhead and requires developers to clone/update repositories for access, which is more cumbersome compared to S3.

The requirement is to automatically delete objects for basic users after 90 days with the least development effort. Amazon S3 provides a built-in, fully managed feature for this: S3 Lifecycle configuration rules . Lifecycle rules can filter objects by prefix (and/or tags) and then perform actions such as expiring (deleting) objects after a specified number of days since creation. Because the bucket does not have versioning enabled, the relevant action is to expire the current object after the retention period.

Option C matches this exactly: create an S3 Lifecycle rule that applies only to objects under the basic/ prefix and configure the rule to expire current versions after 90 days . S3 then automatically and continuously enforces the policy without any code, scheduling, or operational maintenance. This is typically the lowest-effort and most reliable approach because it is handled by S3 itself.

Option A requires writing and operating a Lambda function plus a scheduler, handling pagination, permissions, error handling, retries, and potential costs and throttling when scanning large buckets. That is more development and operational effort than a lifecycle rule. Option B is incorrect because it targets the premium/ prefix, but premium objects must be retained indefinitely. Option D refers to delete markers and unfinished multipart uploads; delete markers are relevant primarily for versioned buckets , and cleaning up multipart uploads does not implement the required “delete basic objects after 90 days” behavior.

Therefore, the correct solution is C : use an S3 Lifecycle rule scoped to the basic/ prefix to expire objects after 90 days, achieving automated retention with minimal development effort.

To rotate database credentials without downtime, Secrets Manager recommends the alternating users (dual user) rotation strategy. With a single-user strategy, Secrets Manager changes the password for the same database user that the application is actively using. If the application continues using cached/old credentials during rotation, authentication failures and downtime can occur.

With the alternating users strategy, two database users exist (often named something like appuser and appuser_clone). Secrets Manager alternates which user is “active” in the secret. During rotation, Secrets Manager updates the inactive user’s password, verifies it, updates the secret to point to the newly updated user, and then (optionally) retires the old credentials. This approach minimizes disruption because there is always one known-good credential set while the other is being updated.

The question specifically requires regular rotation using Secrets Manager. That is achieved by enabling automatic rotation on the secret with a schedule (for example, every 30/60/90 days). Automatic rotation invokes the rotation Lambda (AWS-provided or configured) and performs the workflow without manual intervention.

Option D combines both requirements: automatic rotation plus the alternating users strategy for high availability and no downtime.

Option B mentions alternating users but “managed rotation” is not the standard term used in these choices as the primary differentiator—automatic rotation is the necessary capability to rotate regularly without manual effort.

Options A and C (single user) are more likely to cause downtime during rotation.

Therefore, configure automatic rotation with the alternating users rotation strategy.

AWS IAM role permissions are evaluated dynamically . According to AWS documentation, when an IAM role attached to an EC2 instance is updated, the updated permissions are available to applications running on that instance without requiring a restart . The EC2 instance retrieves temporary credentials through the instance metadata service, and those credentials are refreshed automatically.

In this scenario, the least disruptive approach is to update the IAM role policy to include the required s3:GetObject permission. Once the permission is added, the application can immediately retry access to the S3 bucket without restarting, stopping, or terminating the EC2 instance.

Terminating or restarting the instance (Options A and C) introduces unnecessary downtime and operational overhead. Modifying the S3 bucket policy (Option D) is unnecessary because the problem lies with missing role permissions, not bucket access configuration.

AWS best practices strongly recommend using IAM roles for EC2 and updating permissions dynamically to avoid service interruption. Therefore, simply adding the required permission to the role is the correct and most efficient solution.

Containerization: Packaging the application as a container image promotes portability and standardization. Docker is the standard tool for containerization.

Amazon ECR: ECR is a managed container registry designed to work seamlessly with AWS container services.

Fargate: ECS Fargate provides serverless container orchestration, minimizing operational overhead for this proof-of-concept.

Requirement Summary:

DynamoDB table Orders

Partition key: OrderID

No sort key

100,000+ records

Need to efficiently retrieve all records where:

OrderSource = " MobileApp "

Must optimize for user experience (i.e., performance)

Evaluate Options:

Option A: Scan with FilterExpression

Inefficient for large tables

A Scan reads every item in the table and then applies a filter condition.

This is slow, expensive, and not scalable as the dataset grows.

Option B: Create a Local Secondary Index (LSI) with OrderSource

Invalid: LSIs require the same partition key (OrderID) as the base table.

Since OrderSource is not part of the primary key, and we want to query based on it, LSI won’t help.

Option C: Create a GSI with OrderSource as the sort key

Misuse of sort key: Querying by sort key alone is not allowed in DynamoDB.

You must specify a partition key in a Query.

This design is not valid for the use case.

Option D: Create a GSI with OrderSource as the partition key

BEST CHOICE: With a GSI on OrderSource, you can query efficiently for all items where OrderSource = " MobileApp " .

DynamoDB GSIs allow an alternative access pattern for queries not supported by the base table schema.

GSI Design: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/GSI.html

Querying a GSI: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/Query.html

Scan vs Query: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/bp-query-scan.html

Why Option A is Correct:Auto-scaling with a target utilization ensures the DynamoDB table dynamically adjusts capacity based on workload, maintaining high performance while optimizing cost. Setting a reasonable target utilization minimizes overprovisioning and throttling risks.

Why Other Options are Incorrect:

Option B: On-demand capacity is costlier than provisioned throughput for predictable workloads.

Option C: Using manual CloudWatch alarms and Lambda for scaling is less efficient and adds operational overhead.

Option D: DAX accelerates read performance but does not improve write performance.

AWS Documentation References:

DynamoDB Auto Scaling

Why Option A is Correct:

The ArnLike condition with the specific ARN for myStateMachine ensures that only this state machine can assume the role. The format urn:aws:states is correct for specifying Step Functions resources.

Why Other Options are Incorrect:

Option B: Wildcards (*) in the ARN allow more resources to assume the role, which violates the requirement.

Option C: This condition restricts the account but not the specific state machine.

Option D: A StringNotEquals condition is used to deny specific values, which does not ensure exclusivity for the desired state machine.

AWS Documentation References:

IAM Trust Policies for Step Functions

Comprehensive and Detailed Step-by-Step Explanation:

Option C: Use AWS CodePipeline for CI/CD Workflow:

AWS CodePipeline automates the entire CI/CD pipeline, including build, deploy, and test stages. This minimizes manual effort and integrates well with AWS services.

API Gateway Stages: Represent different environments, such as dev, test, and prod, allowing isolated deployment and testing.

AWS CloudFormation Templates: Ensure that the infrastructure for Lambda and API Gateway is consistent across environments.

AWS CodeBuild for Automated Tests: Validates the deployments in each stage, ensuring integration and functionality are tested post-deployment.

Why Other Options Are Incorrect:

Option A and B: While using AWS SAM or CloudFormation for infrastructure management is valid, these options lack the fully automated CI/CD pipeline provided by CodePipeline.

Option D: Manually invoking Lambda functions using the AWS CLI introduces operational overhead and lacks the automation provided by CodePipeline.

The requirement is specific: certain JSON fields must be encrypted at the edge (before traversing the origin path) and must remain encrypted end-to-end through CloudFront → API Gateway → Lambda → DynamoDB. The AWS feature designed for encrypting specific fields at the edge is CloudFront field-level encryption. With field-level encryption, CloudFront encrypts designated fields in an HTTPS request before forwarding the request to the origin. The origin (API Gateway/Lambda) receives the encrypted values, and only trusted components that have the private key can decrypt them—ensuring the fields remain protected throughout the stack.

Option B matches this exactly: field-level encryption uses an asymmetric key pair (public key used by CloudFront to encrypt; private key held securely by the application for decryption). This satisfies “encrypted at the edge” and “remain encrypted throughout the full stack.”

The “PII encrypted at rest in DynamoDB” requirement is also satisfied because DynamoDB supports encryption at rest by default (and can use AWS owned or customer managed KMS keys). Field-level encryption adds granular protection for only the most sensitive attributes beyond standard at-rest encryption.

Option A (Lambda@Edge encryption) could encrypt at edge, but it is more custom code, higher operational complexity, and is not as direct/standard as CloudFront field-level encryption for encrypting request fields.

Option C encrypts in Lambda, not at the edge—data would already have traversed CloudFront and API Gateway in plaintext form before Lambda encrypts it, violating the “encrypted at the edge” requirement.

Option D is not a defined mechanism for edge encryption; API Gateway methods do not provide edge field encryption with KMS in this way.

Therefore, CloudFront field-level encryption is the correct solution.

AWS CodeArtifact emits events when repository actions occur, such as publishing a new package. AWS documentation recommends Amazon EventBridge as the native service for routing AWS service events to targets with minimal configuration and no infrastructure to manage.

By creating an EventBridge rule that matches CodeArtifact package publication events and routing those events directly to an Amazon SNS topic , the developer can notify the team immediately. SNS provides built-in email subscriptions, retries, and delivery guarantees without custom code.

Option C is the most operationally efficient because it uses fully managed, serverless services and requires no custom compute, code maintenance, or orchestration. EventBridge directly supports SNS as a target, making the configuration straightforward.

Option A is invalid because CodeArtifact does not directly publish notifications to SNS topics. Options B and D introduce unnecessary complexity by adding Lambda or Step Functions and Amazon SES, increasing operational overhead without added value.

Therefore, using EventBridge to route CodeArtifact events to SNS is the AWS-recommended and simplest solution.

Lambda Layers: These are designed to package dependencies that you can share across functions.

How to Use:

Create a layer, upload your 100MB library as a zip.

Attach the layer to your function.

In your function code, import the library from the standard layer path.

The requirement is to search for partial matches of email_address but scoped to a specific customer_type. In DynamoDB, efficient partial matching is performed with a Query on a sort key using the begins_with() condition. However, Query requires that the partition key be specified exactly, so we need an index where customer_type can be used as the partition key, and email_address can be used as the sort key for prefix matching.

A Global Secondary Index (GSI) can be added to an existing table without recreating it, and it can use a different partition key and sort key from the base table. Option A correctly proposes a GSI with:

Partition key: customer_type

Sort key: email_address

Then the Lambda can run a Query on the GSI like: customer_type = :ct AND begins_with(email_address, :prefix) which returns emails that start with the typed prefix for that customer type.

Option B is wrong because it keeps email_address as the partition key. That would require an exact email value (not a prefix) to Query, and you cannot do begins_with on a partition key; begins_with is for sort keys.

Options C and D are invalid because Local Secondary Indexes (LSIs) must share the same partition key as the base table (here, email_address). You cannot create an LSI with customer_type or job_title as the partition key when the table partition key is email_address.

Therefore, adding a GSI (customer_type PK, email_address SK) and querying it with begins_with is the correct solution.

The requirement is to redact PII before the object content reaches the consumer, while the analytics service continues to use standard S3 GET requests. The most operationally efficient way is Amazon S3 Object Lambda.

S3 Object Lambda lets you add your own code to process and transform data as it is retrieved from S3. Instead of the analytics service reading directly from the bucket, it reads through an S3 Object Lambda Access Point. When the service performs a GET request, S3 invokes the associated Lambda function, which can modify the object payload on the fly—for example, by calling an external or internal PII redaction API and returning a redacted version of the report. The original object remains unchanged in the bucket, while consumers receive the transformed (redacted) response.

This approach is operationally efficient because it avoids:

duplicating and storing redacted copies of every report,

refactoring the analytics service to use a different datastore or ingestion path,

building a separate proxy service layer.

Option A requires significant refactoring and ongoing ETL/warehouse ops.

Option C provides confidentiality via encryption but does not perform redaction (analytics would still see PII after decryption).

Option D changes the access pattern completely (analytics no longer uses GET) and adds messaging complexity without directly returning redacted object content.

Therefore, S3 Object Lambda + Object Lambda Access Point is the most efficient solution to redact data at retrieval time.

To ensure a Lambda function URL is reachable only through CloudFront, the solution must (1) make CloudFront the only authorized caller, and (2) prevent direct public access to the function URL endpoint. The AWS-native way to do this is to use a resource-based policy on the Lambda function that permits invocation only from the specific CloudFront distribution, combined with CloudFront’s Origin Access Control (OAC) to sign origin requests.

A Lambda function URL can be configured for IAM-based authorization, and Lambda supports resource-based permissions (via lambda:AddPermission) that restrict who can invoke it. By scoping the permission so that only the CloudFront distribution (identified by a source ARN / distribution identifier) is allowed, direct requests that do not come through CloudFront will be rejected. This enforces the requirement that the function URL cannot be accessed directly.

CloudFront OAC is the modern mechanism to securely access origins by having CloudFront sign the requests it sends to the origin. When CloudFront signs requests and the origin (here, the Lambda function URL) is configured to accept only authorized requests per its resource policy, CloudFront becomes the only viable access path.

Option A is incorrect because CloudFront does not use a “resource-based policy” to grant access to an origin in this way; the enforcement must happen at the origin. Option C is not how CloudFront accesses origins; CloudFront does not assume a customer IAM role to fetch origin content. Option D is fragile and not recommended: CloudFront IP ranges can change and are not intended to be used as a static allowlist for origin protection; additionally, IP-based controls do not provide the same strong identity-based authorization that OAC + resource policy provides.

Therefore, B is the correct solution: restrict the Lambda function URL with a Lambda resource-based policy and configure CloudFront with OAC so only CloudFront-originated requests are accepted.

The solution that will meet the requirements with the least development effort is to set the image resize Lambda function as a destination of the avatar generator Lambda function for the events that fail processing. This way, the fallback mechanism is automatically triggered by the Lambda service without requiring any additional components or configuration. The other options involve creating and managing additional resources such as queues, topics, state machines, or rules, which would increase the complexity and cost of the solution.

Amazon API Gateway stages are designed to represent different deployment environments such as development, testing, and production. AWS documentation recommends using separate stages combined with stage variables to reference different backend resources without duplicating API definitions.

In this scenario, the development stage already uses a stage variable to reference the Lambda dev alias. To expose the production alias safely, the developer should deploy the API to a new stage named production and configure a stage variable that points to the Lambda prod alias. This approach allows the same API configuration and methods to be reused while cleanly separating environments.

Stage variables are resolved at runtime and are commonly used to reference Lambda aliases in integration ARNs. This enables controlled promotion of code from development to production without modifying methods or integrations.

Creating new methods (Options A and B) is unnecessary and increases operational complexity. Directly hardcoding the Lambda alias in the integration (Option D) removes flexibility and deviates from AWS best practices for environment separation.

Therefore, deploying a new API stage and using stage variables to reference the prod Lambda alias is the correct and most efficient solution.

Amazon CloudFront is a global content delivery network (CDN) designed to deliver static and streaming content with low latency and high throughput . AWS documentation explicitly recommends CloudFront for media streaming and high-request-rate workloads.

By placing CloudFront in front of Amazon S3, video content is cached at edge locations close to end users. This dramatically reduces latency, minimizes buffering, and offloads request volume from the S3 bucket. CloudFront is built to handle sudden traffic spikes and supports tens of thousands of requests per second per edge location.

Route 53 routing policies (Option A) control DNS resolution but do not cache content. Cross-Region Replication (Option B) improves regional availability but does not provide edge caching or streaming optimization. S3 Intelligent-Tiering (Option C) optimizes storage cost, not performance.

Therefore, using CloudFront with S3 as the origin provides the largest and most immediate performance improvement.

To test an updated Lambda integration without impacting production traffic, the key is to isolate test requests from the production stage that real users call. In API Gateway, a stage represents a deployed snapshot of an API configuration, and each stage has its own invoke URL . Creating a separate test stage provides immediate traffic isolation because production users continue using the production stage URL, while the developer tests using the test stage URL.

Using stage variables is a common API Gateway pattern to parameterize integrations per stage. For Lambda backends, stage variables can reference different Lambda function ARNs, aliases, or versions (depending on the integration style), allowing the same API resources and methods to route to different backends per stage. This lets the developer deploy and validate the updated Lambda function only in the test stage while keeping the production stage pointing to the stable Lambda version. The result is operationally efficient: minimal duplicated infrastructure, clear separation, and easy rollback by switching stage variables or redeploying only the stage configuration.

Option A (canary on the existing stage) intentionally shifts a percentage of production traffic to the new version, which violates the requirement that testing must not affect any production users. Option B (private endpoint) changes network accessibility and can break production access; it also does not provide safe version testing. Option D (clone entire stack) can work but has higher operational overhead (duplicate resources, additional deployment management) than using native stage isolation.

Therefore, C best meets the requirements: create a new test stage , use stage variables to point that stage to the updated Lambda function, and test safely using the test stage URL while production remains unaffected.

Why Option B is Correct:RDS Proxy manages database connections efficiently, reducing overhead on the RDS instance and mitigating " too many connections " errors.

Why Other Options are Incorrect:

Option A: DAX is for DynamoDB, not RDS.

Option C: Migration to PostgreSQL does not address the current issue.

Option D: ElastiCache is useful for caching but does not solve connection pool issues.

AWS Documentation References:

Amazon RDS Proxy

The most secure solution is A because AWS Secrets Manager is purpose-built for storing, retrieving, and managing secrets such as database credentials, API keys, and tokens. In AWS guidance, Secrets Manager is recommended when applications need secure secret storage with controlled access, encryption, and integration with runtime retrieval patterns. By storing the credentials as a secret and passing only the secret ARN through an environment variable, the Lambda function avoids exposing the actual username and password in plaintext configuration.

The AWS Parameters and Secrets Lambda Extension is specifically designed to let Lambda functions retrieve secrets securely at invocation time. This reduces the risk of accidental exposure in code, deployment artifacts, or function configuration. Secrets Manager also supports automatic rotation , which is a major security advantage for database credentials and is one of the main reasons AWS recommends it over hardcoded or manually managed values.

Option B is incorrect because base64 is only encoding, not encryption, and embedding credentials in source code is insecure. Option C is less appropriate because a string type parameter in Systems Manager Parameter Store is not the secure choice for sensitive credentials; secure storage there would require SecureString , but even then Secrets Manager is the stronger AWS-recommended service for database secrets. Option D only masks values in CloudFormation outputs and logs with NoEcho ; it does not provide secure runtime secret retrieval or secret lifecycle management.

Therefore, A is the best and most secure answer according to AWS security best practices for Lambda applications handling database credentials.

Using a dead-letter queue (DLQ) with Amazon SQS is the most operationally efficient solution for handling unprocessable messages.

Amazon SQS Dead-Letter Queue:

A DLQ is used to capture messages that fail processing after a specified number of attempts.

Allows the application to continue processing other messages without being blocked.

Messages in the DLQ can be analyzed later for debugging and resolution.

Why DLQ is the Best Option:

Operational Efficiency: Automatically defers messages with errors, ensuring the queue is not blocked.

Analysis Ready: Messages in the DLQ can be inspected to identify recurring issues.

Scalable: Works seamlessly with Lambda and SQS at scale.

Why Not Other Options:

Option A: Logs the messages but does not resolve the queue blockage issue.

Option C: FIFO queues and 0-second retention do not provide error handling or analysis capabilities.

Option D: Alerts administrators but does not handle or store the unprocessable messages.

Steps to Implement:

Create a new SQS queue to serve as the DLQ.

Attach the DLQ to the primary queue and configure the Maximum Receives setting.

Using Amazon SQS Dead-Letter Queues

Best Practices for Using Amazon SQS with AWS Lambda

This is a cross-account access pattern using STS AssumeRole. Account A owns the DynamoDB table and has created an IAM role (AccessPII) with permissions to access the table. Account A also configured a trust policy that allows principals from Account B to assume the role. That trust policy alone is not enough; the caller in Account B must also be allowed to call sts:AssumeRole.

Therefore, in Account B, the EC2 instance profile role (the role attached to the EC2 instances running the app) must have permission to assume the role in Account A. That is Option A.

Next, the application must actually obtain temporary credentials for the Account A role. The standard way is to call STS AssumeRole in the application logic (or use an SDK credential provider that assumes a role). This yields temporary credentials that have the permissions of the AccessPII role, allowing DynamoDB access in Account A. That is Option D.

Option B is not correct because granting direct DynamoDB table permissions in Account B does not grant access to a table owned by Account A unless Account A also grants access via a resource policy (DynamoDB resource policies exist but the scenario already sets up role assumption; the intended solution is AssumeRole).

Option C is incomplete/wrong: getting temporary credentials from the EC2 role alone gives permissions of the EC2 role in Account B, not the permissions in Account A. You must assume the cross-account role.

Option E is for IAM user MFA session tokens, not cross-account role assumption for an EC2 role.

Therefore, grant sts:AssumeRole to the EC2 role and call AssumeRole in code.

Comprehensive and Detailed 250 to 300 words of Explanation (AWS-doc-aligned):

The most operationally efficient way to share and patch common dependencies across multiple Lambda functions is to use AWS Lambda layers . A Lambda layer is an independently managed deployment artifact that can contain libraries and other dependencies. Multiple functions can reference the same layer, which avoids packaging identical open source modules into each function bundle.

When a vulnerability or bug is found in a shared module, you patch it once by publishing a new layer version that contains the updated dependencies. Then you update each Lambda function configuration to reference the new layer version. This centralizes dependency management, reduces duplicated artifacts, and makes rollouts and rollbacks cleaner (you can switch layer versions quickly). It also keeps function deployment packages smaller, which often improves deployment speed and can reduce cold-start overhead associated with larger bundles.

The other options introduce unnecessary operational burden:

Downloading modules at cold start (CloudFront/S3) increases latency and adds failure modes at runtime.

Hosting a private registry on EC2 requires patching, scaling, availability management, and security hardening—high operational overhead.

A CloudFormation registry extension and daily scans are not a standard or efficient mechanism for packaging runtime Node.js dependencies into Lambda functions.

Lambda layers are a documented AWS-native feature intended for precisely this use case: sharing common code and libraries across functions while enabling controlled versioning. Therefore, creating a layer for the open source modules and updating the functions to consume the layer is the most efficient and maintainable approach.

Lambda Destinations on Failure: Allow routing asynchronous function invocations to specified resources (like another Lambda function) upon failure.

Error Handling: The error-handling Lambda receives details about the failure, enabling logging and custom actions.

Direct Integration: This solution leverages native Lambda functionality for a simpler implementation.

Comprehensive and Detailed Explanation (AWS Verified – 250–300 Words)

When an Amazon S3 bucket is configured with default encryption using SSE-KMS, AWS Key Management Service (KMS) plays an active role in every PutObject request. Although the IAM user already has the s3:PutObject permission, this alone is not sufficient when SSE-KMS is enabled.

According to AWS documentation, when a client uploads an object to an S3 bucket encrypted with SSE-KMS, Amazon S3 must call AWS KMS on behalf of the caller to generate a unique data encryption key. This operation requires the kms:GenerateDataKey permission on the KMS key used for encryption. If this permission is missing, the PutObject request fails with an Access Denied error—even though S3 permissions appear correct.

AWS explicitly states that IAM principals uploading objects to SSE-KMS–encrypted buckets must be allowed to use the KMS key. At a minimum, the following KMS permissions are required:

kms:GenerateDataKey

(Implicitly) access allowed by the KMS key policy

Option C correctly resolves the issue by granting the IAM user permission to generate a data key via AWS KMS, enabling successful encryption during the upload process.

Option A is invalid because s3:EncryptionConfiguration is not required for object uploads.

Option B is unnecessary because the IAM user already has s3:PutObject.

Option D is incorrect because ACLs are not used to control KMS encryption permissions and are discouraged by AWS best practices.

Why Option A is Correct:DynamoDB supports incremental exports to Amazon S3 natively, making data analytics-ready with minimal operational overhead.

Why Other Options are Incorrect:

Option B: DynamoDB Streams require additional processing logic to write to S3, increasing complexity.

Option C & D: Using EMR for data movement adds unnecessary operational overhead compared to native exports.

AWS Documentation References:

DynamoDB Exports to S3

Comprehensive and Detailed Step-by-Step Explanation:

Issue Analysis:

The stream uses service name as the partition key. This can cause " hot partition " issues when a few service names generate significantly more logs compared to others, causing uneven distribution of data across shards.

Metrics show that the write capacity used is below provisioned capacity, which confirms that the throughput errors are due to shard-level limits and not overall capacity.

Option C: Change Partition Key to Creation Timestamp:

By changing the partition key to the creation timestamp (or a composite key including timestamp), the distribution of data across shards can be randomized, ensuring an even spread of records.

This resolves the shard overutilization issue and eliminates ProvisionedThroughputExceededException.

Why Other Options Are Incorrect:

Option A: Switching to on-demand capacity mode might temporarily alleviate the issue, but the root cause (hot partitioning) remains unresolved.

Option B: Adding shards increases capacity but does not fix the skewed data distribution caused by using the service name as the partition key.

Option D: Creating separate streams for each service adds unnecessary complexity and does not scale well as the number of services grows.

The company already has SSM Agent deployed on all servers, which enables centralized, scalable fleet operations without logging in to each instance. The most operationally efficient solution is to automate key distribution using Systems Manager Run Command, executing a script across the entire fleet (or a targeted subset) in one action.

Option B uses S3 as a simple distribution source and Run Command to perform the update at scale. The developer can store the required key material in a controlled S3 location and write a script that places the key in the correct filesystem path (with correct permissions/ownership), updates authorized_keys if needed, and restarts any services if required. Run Command can target instances by tags, resource groups, or instance IDs and provides execution logging and status.

Option A is not operationally efficient because it requires logging into each server manually.

Option C and D push the distribution burden to humans (“grant team members permissions to download”), which is error-prone, slow, and not scalable. Also, distributing private keys broadly to individuals increases risk.

Comprehensive Detailed and Lengthy Step-by-Step Explanation with All AWS Developer References:

1. Understanding the Use Case:

The developer needs to export DynamoDB table data into Amazon S3 buckets using the AWS CLI, and some exports are failing. Proper credentials and permissions have already been configured.

2. Key Conditions to Check:

Region Consistency:DynamoDB exports require that the target S3 bucket and the DynamoDB table reside in the same AWS Region. If they are not in the same Region, the export process will fail.

Point-in-Time Recovery (PITR):PITR is not required for exporting data from DynamoDB to S3. Enabling PITR allows recovery of table states at specific points in time but does not directly influence export functionality.

DynamoDB Streams:Streams allow real-time capture of data modifications but are unrelated to the bulk export feature.

DAX (DynamoDB Accelerator):DAX is a caching service that speeds up read operations for DynamoDB but does not affect the export functionality.

3. Explanation of the Options:

Option A: " Ensure that point-in-time recovery is enabled on the DynamoDB tables. " While PITR is useful for disaster recovery and restoring table states, it is not required for exporting data to S3. This option does not address the export failure.

Option B: " Ensure that the target S3 bucket is in the same AWS Region as the DynamoDB table. " This is the correct answer. DynamoDB export functionality requires the target S3 bucket to reside in the same AWS Region as the DynamoDB table. If the S3 bucket is in a different Region, the export will fail.

Option C: " Ensure that DynamoDB streaming is enabled for the tables. " Streams are useful for capturing real-time changes in DynamoDB tables but are unrelated to the export functionality. This option does not resolve the issue.

Option D: " Ensure that DynamoDB Accelerator (DAX) is enabled. " DAX accelerates read operations but does not influence the export functionality. This option is irrelevant to the issue.

4. Resolution Steps:

To ensure successful exports:

Verify the Region of the DynamoDB tables:

Check the Region where each table is located.

Verify the Region of the target S3 buckets:

Confirm that the target S3 bucket for each export is in the same Region as the corresponding DynamoDB table.

If necessary, create new S3 buckets in the appropriate Regions.

Run the export command again with the correct setup:

aws dynamodb export-table-to-point-in-time \

--table-name < TableName > \

--s3-bucket < BucketName > \

--s3-prefix < Prefix > \

--export-time < ExportTime > \

--region < Region >

The goal is to reduce a given item’s price by 100 only when the resulting value would not fall below 500 . That means the update should occur only when the current Price is at least 600 (because Price - 100 > = 500 ⇒ Price > = 600). The requirement also says to do this with the least number of calls to DynamoDB. The most efficient approach is to use a single UpdateItem call with a ConditionExpression so DynamoDB enforces the rule atomically.

Option B does exactly that: issue one UpdateItem request that updates Price = Price - :delta (using an update expression) while including a condition like Price > = :minPriceBeforeReduction (where :minPriceBeforeReduction is 600). If the condition is true, DynamoDB performs the update; if it’s false, DynamoDB rejects the write with ConditionalCheckFailedException. This avoids a separate read/query call and prevents race conditions where the price could change between a read and a subsequent write.

Option A requires two calls (Query then UpdateItem) and is vulnerable to time-of-check/time-of-use issues unless additional conditional logic is added to the update anyway. Option C adds an extra condition on category, but the requirement is “reduce all inventory prices” subject to the minimum resulting price; filtering by category is unnecessary and can cause eligible items in other categories to be skipped. Option D expresses the rule in a mathematical form; while conceptually correct, DynamoDB condition expressions do not generally support arbitrary arithmetic in the condition the same way as a simple comparison, so the standard implementation is to compare against 600 directly.

Therefore, B is the correct choice: one UpdateItem call with a condition expression Price > = 600 minimizes DynamoDB calls and enforces the pricing rule safely and atomically.

Maximum concurrency for SQS as an event source allows customers to control the maximum concurrent invokes by the SQS event source1. When multiple SQS event sources are configured to a function, customers can control the maximum concurrent invokes of individual SQS event source1.

In this scenario, the developer needs to resolve the issue of the third-party API frequently returning an HTTP 429 Too Many Requests error message, which prevents a significant number of messages from being processed successfully. To achieve this, the developer can follow these steps:

Find out the documented rate limits of the third-party API, which specify how many requests can be made in a given time period.

Configure maximum concurrency on the SQS event source based on the rate limits of the third-party API. This will limit the number of concurrent invokes by the SQS event source and prevent exceeding the rate limits of the third-party API.

Test and monitor the application performance and adjust the maximum concurrency value as needed.

By using this solution, the developer can reduce the frequency of HTTP 429 errors and improve the message processing success rate. The developer can also avoid throttling or blocking by the third-party API.

The correct answer is B because Amazon CloudWatch Logs Insights is the AWS-native tool for interactively searching, filtering, and analyzing log data that is already stored in CloudWatch Logs. In this case, the developer needs to investigate intermittent Lambda failures within a specific 1-hour period , identify logs related to failures, and capture the timestamps of those failures. Logs Insights is purpose-built for exactly this kind of operational troubleshooting.

AWS documentation describes Logs Insights as a fully integrated log analytics capability that can query log groups by time range and pattern. The developer can run a query that filters for error-related log messages, extracts timestamp fields, and sorts the results to find exactly when failures occurred. This is highly efficient because it does not require exporting logs, writing custom parsing scripts, or manually browsing through streams.

Option A is possible but not the most operationally efficient. Browsing logs through the CLI is slower and more manual, especially for intermittent failures. Option C is inefficient because downloading logs to a local machine adds unnecessary steps and operational overhead. Option D can work for large-scale analytics, but exporting to S3 and querying with Athena is more complex than needed for logs that are already in CloudWatch Logs.

Therefore, the best solution is to use CloudWatch Logs Insights to query the Lambda log group for failure patterns and timestamps. This provides fast, targeted, and operationally efficient troubleshooting directly within AWS.

Amazon SQS standard queues provide at-least-once delivery, which means messages can be delivered more than once. When Lambda is triggered by an SQS standard queue, duplicate delivery can occur due to retries, visibility timeouts, or transient errors. Therefore, “processed multiple times” is expected behavior unless the system implements deduplication or idempotency.

The most cost-effective way within the provided options to reduce duplicate processing is to use an SQS FIFO queue with deduplication. FIFO queues support exactly-once processing semantics within the constraints of the service (by preventing duplicate message delivery within the deduplication interval when a MessageDeduplicationId is used or content-based deduplication is enabled). This directly mitigates duplicate processing without requiring large architectural changes.

Option B (DLQ) is important for handling poison messages, but it does not prevent duplicates in normal processing; it only captures messages that fail repeatedly.

Option C (concurrency = 1) reduces parallelism but does not eliminate duplicates; the same message can still be delivered again if the visibility timeout expires or retries occur.

Option D is a major redesign and not cost-effective for simply addressing duplicate SQS message processing.

???? Note: In real production systems, the best practice is to make processing idempotent (so duplicates do no harm). But among the choices, FIFO + deduplication is the most direct and cost-effective fix.

Therefore, switch to an SQS FIFO queue and use deduplication.

Pseudo Parameters: CloudFormation provides pseudo parameters that reference runtime context, including the current AWS Region.

Operational Efficiency: The AWS::Region pseudo parameter offers the most direct and self-contained way to obtain the Region dynamically within the template.

CloudFormation and Dynamic References: The DefinitionSubstitutions property in CloudFormation allows you to pass values into Step Functions state machines at runtime.

Cost-Effectiveness: This solution is cost-effective as it leverages CloudFormation ' s built-in capabilities, avoiding the need for additional services like Secrets Manager or AppConfig.

https://docs.aws.amazon.com/IAM/latest/UserGuide/reference_policies_examples_lambda-access-dynamodb.html

The solution that will solve this problem is to create and inspect the Lambda dead-letter queue. Troubleshoot the failed functions. Reprocess the events. This way, the developer can identify and fix any issues that caused the Lambda function to fail when invoked asynchronously by API Gateway. The developer can also reprocess any orders that were not processed due to failures. The other options either do not address the root cause of the problem, or do not help recover from failures.

Comprehensive and Detailed Explanation (250–300 words):

AWS best practices for multi-tenant isolation recommend session-based access control using IAM session tags . Session tags allow temporary credentials to be scoped dynamically based on tenant context.

First, the data access role must allow sts:TagSession (Option A) so tenant identifiers can be passed securely at runtime. The Lambda function assumes this role and includes the tenant name as a session tag (Option F).

Next, IAM policies on the data access role restrict access to S3 object prefixes and DynamoDB partition keys by evaluating the session tag (Option C). This ensures that even if a bug occurs, the role cannot access data belonging to another tenant.

Resource-based DynamoDB policies and RCPs are unnecessary here and add complexity.

This approach follows AWS’s recommended attribute-based access control (ABAC) model and provides strong tenant isolation with minimal operational overhead.

This solution allows the developer to keep the dependencies of the Lambda functions up to date with the least additional complexity because it eliminates the need to update each function individually. A Lambda layer is a ZIP archive that contains libraries, custom runtimes, or other dependencies. The developer can create a layer that contains all of the shared dependencies and attach it to multiple Lambda functions. When the developer updates the layer, all of the functions that use the layer will have access to the latest version of the dependencies.

With Lambda’s SQS event source mapping, messages are delivered with at-least-once semantics. When processing messages in batches (up to 10 per invocation), a common source of unnecessary downstream API calls occurs when only some records in the batch fail. Without special handling, a failure can cause the entire batch to be retried, which means records that were already successfully processed can be delivered again, leading to duplicate calls to the third-party API.

To minimize unnecessary API calls while still preserving at-least-once delivery, the best solution is to enable partial batch response (also described as “report batch item failures”). When enabled, the Lambda function returns a response that explicitly identifies which message(s) failed (by messageId). The event source mapping then retries only the failed messages , while successfully processed messages are deleted from the queue. This reduces duplicates and avoids re-sending already successful records to the third-party API, while still ensuring failed items are retried until they succeed or are moved to a DLQ (if configured).

Option A (FIFO + content-based deduplication) helps prevent duplicate enqueue events within the deduplication window, but it does not stop duplicate deliveries caused by retries, visibility timeouts, or partial failures in a batch. Option B (reserved concurrency = 1) reduces parallelism but does not prevent retries or batch-level duplication; it can also hurt throughput. Option D (DLQ) is important for poison-pill handling but does not reduce duplicate API calls during normal retry behavior; it merely captures messages after the maximum receive count.

Therefore, C is the correct choice: enable batch item failure reporting and implement the function to return partial batch responses so only failed messages are retried, minimizing unnecessary third-party API calls while maintaining at-least-once delivery.

Why Option B is Correct:A customer-managed AWS Key Management Service (KMS) key allows for encryption at rest and provides the ability to rotate the key on demand. This ensures compliance with security requirements for key management and database encryption.

RDS integrates natively with AWS KMS, allowing the use of a customer-managed key for encrypting data at rest.

Key rotation can be managed directly in AWS KMS without needing custom solutions.

Why Other Options are Incorrect:

Option A: AWS KMS managed encryption keys (AWS-owned keys) do not support key rotation on demand.

Option C & D: Storing keys in AWS Secrets Manager with custom rotation is not a recommended approach for database encryption. AWS KMS is designed specifically for secure key management and encryption.

AWS Documentation References:

Encrypting Amazon RDS Resources

AWS Key Management Service (KMS)

Principle of Least Privilege: Granting specific permissions through an IAM role is more secure than directly attaching policies to a function or using root user credentials.

IAM Roles for Lambda: Designed to provide temporary credentials to Lambda functions, enhancing security.

Reusability: The existing IAM policy ensures the correct S3 and DynamoDB access is granted.

Requirement Summary:

Simple REST endpoint for weather data

Light backend usage (POC, testing)

Wants caching support to reduce backend calls

Must be cost-effective

Evaluate Options:

B: Lambda + API Gateway + SAM

Serverless = No idle costs

API Gateway can enable caching (response caching at endpoint level)

SAM makes deployment simple and repeatable

Perfect for low-traffic testing

A: EKS + API Gateway

High overhead

Not cost-effective for POC/testing

C: ECS + API Gateway

Similar to A: Container orchestration not needed for light REST endpoint

D: Elastic Beanstalk + ALB + Lambda

Overly complex and does not directly expose Lambda

Beanstalk better suited for full apps, not small REST functions

 AWS SAM: https://docs.aws.amazon.com/serverless-application-model/latest/developerguide/what-is-sam.html

 API Gateway caching: https://docs.aws.amazon.com/apigateway/latest/developerguide/api-gateway-caching.html

 Serverless Best Practices: https://docs.aws.amazon.com/lambda/latest/dg/best-practices.html

The correct answers are B and E .

The existing platform already runs on Kubernetes , so the fastest and most compatible migration path is to use Amazon EKS , which is AWS’s managed Kubernetes service. This preserves the application’s orchestration model and minimizes refactoring. Because the clusters must be highly available , EKS is a strong fit since it is designed to run Kubernetes workloads across multiple Availability Zones.

For shared storage, the on-premises application currently depends on a common NFS share that all containers can access. On AWS, the service that most closely matches this requirement is Amazon EFS , which provides a managed, elastic, shared file system that supports NFS access and can be mounted concurrently from multiple compute nodes. By contrast, Amazon EBS is block storage that is generally attached to a single instance and is not the right replacement for a shared multi-node NFS workload.

Therefore, B is correct because it moves the shared NFS data to Amazon EFS and stores the container images in Amazon ECR , which is the proper AWS managed registry for container images. E is correct because it runs the applications on Amazon EKS and uses Amazon EFS as the shared mounted file system across the Kubernetes worker nodes.

Option C is wrong because it uses ECS , not Kubernetes. Option D is wrong because EBS does not meet the requirement for a shared NFS-like file system across highly available Kubernetes nodes. Option A is wrong for the same storage reason.

So, the best migration combination is Amazon EFS + Amazon EKS , which makes B and E correct.

Step Functions Retriers: Retriers provide a built-in way to gracefully handle transient errors within State Machines. Here ' s how to use them:

Directly attach a retrier to the problematic ' GetResource ' task.

Configure the retrier:

ErrorEquals: Set this to [ ' TooManyRequestsException ' ] to target the specific error.

IntervalSeconds: Set to 10 for the desired retry delay.

MaxAttempts: Set to 1, as you want only one retry attempt.

Error Handling:

Upon ' TooManyRequestsException ' , the retrier triggers the task again after 10 seconds.

On a second failure, Step Functions moves to the next state or fails the workflow, as per your design.

' IllegalArgumentException ' causes error propagation as intended.

Comprehensive Detailed and Lengthy Step-by-Step Explanation with All AWS Developer References:

1. Understanding the Use Case:

The company wants to gradually release a new feature to 15% of users to perform testing. AWS AppConfig is designed to manage and deploy configurations, including feature flags, allowing controlled rollouts.

2. Key AWS AppConfig Features:

Feature Flags: Enable or disable features dynamically without redeploying code.

Variants: Define different configurations for subsets of users.

Targeting Rules: Specify rules for which users receive a particular variant.

3. Explanation of the Options:

Option A: " Set up a custom script within the application to randomly select 15% of users. Assign a flag for the new feature to the selected users. " While possible, this approach requires significant operational effort to manage user selection and ensure randomness. It does not leverage AWS AppConfig ' s built-in capabilities, which increases overhead.

Option B: " Create separate AWS AppConfig feature flags for both groups of users. Configure the flags to target 15% of users. " Creating multiple feature flags for different user groups complicates configuration management and does not optimize the use of AWS AppConfig.

Option C: " Create an AWS AppConfig feature flag. Define a variant for the new feature, and create a rule to target 15% of users. " This is the correct solution. Using AWS AppConfig feature flags with variants and targeting rules is the most efficient approach. It minimizes operational overhead by leveraging AWS AppConfig ' s built-in targeting and rollout capabilities.

Option D: " Use AWS AppConfig to create a feature flag without variants. Implement a custom traffic splitting mechanism in the application code. " This approach requires custom implementation within the application code, increasing complexity and operational effort.

4. Implementation Steps for Option C:

Set Up AWS AppConfig:

Open the AWS Systems Manager Console.

Navigate to AppConfig.

Create a Feature Flag:

Define a new configuration for the feature flag.

Add variants (e.g., " enabled " for the new feature and " disabled " for no change).

Define a Targeting Rule:

Use percentage-based targeting to define a rule that applies the " enabled " variant to 15% of users.

Targeting rules can use attributes like user IDs or geographic locations.

Deploy the Configuration:

Deploy the configuration using a controlled rollout to ensure gradual exposure.

This solution meets the requirements in the most operationally efficient manner because it does not require any infrastructure provisioning or management. The developer can create a Lambda function that makes the API call and configure an EventBridge rule that triggers the function once a day at a designated time. This is a serverless solution that scales automatically and only charges for the execution time of the function.

The solution that will meet the requirements is to write data to Amazon ElastiCache. This way, the application can write session data to a fast, scalable, and reliable in-memory data store that can be served reliably across multiple requests. The other options either involve writing data to persistent storage, which is slower and more expensive than in-memory storage, or writing data to the root filesystem, which is not shared among multiple EC2 instances.

The amount of memory you allocate to your Lambda function also determines how much CPU and network bandwidth it gets. Increasing the memory size can improve the performance of CPU-bound functions by giving them more CPU power. The CPU allocation is proportional to the memory allocation, so a function with 1 GB of memory has twice the CPU power of a function with 512 MB of memory. Reference: AWS Lambda execution environment

For in-place deployments, AWS CodeDeploy uses a set of predefined hooks that run in a specific order during each deployment lifecycle event. The hooks are ApplicationStop, BeforeInstall, AfterInstall, ApplicationStart, and ValidateService. The run order of the hooks for in-place deployments is as follows:

ApplicationStop: This hook runs first on all instances and stops the current application that is running on the instances.

BeforeInstall: This hook runs after ApplicationStop on all instances and performs any tasks required before installing the new application revision.

AfterInstall: This hook runs after BeforeInstall on all instances and performs any tasks required after installing the new application revision.

ApplicationStart: This hook runs after AfterInstall on all instances and starts the new application that has been installed on the instances.

ValidateService: This hook runs last on all instances and verifies that the new application is running properly on the instances.

The requirement emphasizes collaboration , frequent code changes , and deploying from source control with the fewest manual steps. This aligns with a CI/CD pipeline approach where commits or merges automatically trigger builds and deployments. Using AWS CodePipeline orchestrates the stages (source, build, deploy), and AWS CodeBuild performs the build/package steps for the Lambda/serverless application. When integrated with a repository (such as CodeCommit, GitHub, or another supported provider), CodePipeline can automatically start when changes are merged into specific branches, enabling consistent, repeatable deployments.

Option C provides the least operational overhead for teams: developers push code and merge pull requests; the pipeline automatically builds artifacts, runs tests (if configured), packages dependencies, and deploys updates to Lambda (often via AWS SAM/CloudFormation under the hood). This removes manual “build on my laptop” drift, prevents inconsistent deployments between developers, and supports frequent iteration safely.

Option B (SAM CLI from a local machine) requires a developer to manually run build/deploy commands and assumes each developer’s environment is configured correctly. That increases manual steps and the risk of differences across machines. Option D (uploading zip in the console) is highly manual and not suitable for frequent changes or team collaboration. Option A is also suboptimal because storing built artifacts in source control is an anti-pattern; builds should be reproducible and produced by CI, not committed binaries.

Therefore, C is the best choice: set up CodePipeline + CodeBuild so merges to controlled branches automatically trigger builds and deployments to AWS Lambda with minimal manual intervention.

The correct answer is C. The developer did not update the Docker image tag to a new version.

C. The developer did not update the Docker image tag to a new version. This is correct. When deploying an application to Amazon EKS, the developer needs to specify the Docker image tag that contains the application code and configuration. If the developer does not update the Docker image tag to a new version after making changes to the application, the EKS cluster will continue to use the old Docker image tag that references the original backend resources. To fix this issue, the developer should update the Docker image tag to a new version and redeploy the application to the EKS cluster.

A. The developer did not successfully create the new AWS account. This is incorrect. The creation of a new AWS account is not related to the application’s connection to the backend resources. The developer can use any AWS account to host the EKS cluster and the backend resources, as long as they have the proper permissions and configurations.

B. The developer added a new tag to the Docker image. This is incorrect. Adding a new tag to the Docker image is not enough to deploy the changes to the application. The developer also needs to update the Docker image tag in the EKS cluster configuration, so that the EKS cluster can pull and run the new Docker image.

D. The developer pushed the changes to a new Docker image tag. This is incorrect. Pushing the changes to a new Docker image tag is not enough to deploy the changes to the application. The developer also needs to update the Docker image tag in the EKS cluster configuration, so that the EKS cluster can pull and run the new Docker image.

To meet the requirement that developers cannot view credit card numbers while the fraud detection account can , the solution must (1) ensure the sensitive data is masked at ingestion/storage in CloudWatch Logs, and (2) allow only an authorized principal to view unmasked values .

First, the developers must attach the data protection policy to the CloudWatch log group that receives the API Gateway HTTP API access logs (and/or the Lambda log group if credit card numbers can appear there). A CloudWatch Logs data protection policy is applied at the log group level and instructs CloudWatch Logs to identify sensitive data using managed data identifiers (for example, CreditCardNumber ) and then apply the configured de-identification action (here, masking). Without attaching the policy to the log group, the masking/redaction behavior will not be enforced for newly ingested log events, and developers could still see raw request parameters.

Second, to permit only the fraud detection account to reveal the masked values when necessary, the IAM role that the fraud detection account assumes must have the specific CloudWatch Logs permission to unmask protected data. Granting logs:Unmask to that role enables authorized access to view the sensitive values while keeping them masked for everyone else who lacks that permission (such as developer accounts assuming other roles).

Options A and E are not required for CloudWatch Logs data protection enforcement; the core controls are log-group policy attachment (to perform masking) and IAM authorization (to permit unmasking only for the fraud role). Therefore, the correct combination is C (apply the policy to the log group that captures the HTTP API logs) and B (grant logs:Unmask to the fraud detection role).

This solution meets the requirements in the most secure way because it uses a service that is integrated with CloudFormation to manage sensitive data in encrypted form. AWS Systems Manager Parameter Store provides secure, hierarchical storage for configuration data management and secrets management. You can store sensitive data as secure string parameters, which are encrypted using an AWS Key Management Service (AWS KMS) key of your choice. You can also use dynamic references in your CloudFormation templates to specify template values that are stored in Parameter Store or Secrets Manager without having to include them in your templates. Dynamic references are resolved only during stack creation or update operations, which reduces exposure risks for sensitive data. Putting sensitive data into a CloudFormation parameter will not encrypt them or protect them from unauthorized access. Putting sensitive data into an Amazon S3 bucket or Amazon Elastic File System (Amazon EFS) will require additional configuration and integration with CloudFormation and may not provide fine-grained access control or encryption for sensitive data.

This solution allows the developer to troubleshoot the failure by capturing unprocessed events in a queue for further analysis. Dead Letter Queues (DLQs) are queues that store messages that could not be processed by a service, such as Lambda, for various reasons, such as configuration errors, throttling limits, or permissions issues. The developer can configure DLQs for Lambda functions by sending events to either an Amazon Simple Queue Service (SQS) queue or an Amazon Simple Notification Service (SNS) topic. The developer can then inspect the messages in the queue or topic to identify and fix the root cause of the failure. Configuring AWS CloudTrail logging will not capture invocation failures for asynchronous Lambda invocations, but only record API calls made by or on behalf of Lambda. Configuring Amazon Simple Workflow Service (SWF) or AWS Config will not process any direct unprocessed events, but require additional integration and configuration.

Requirement Summary:

EKS containers send logs to CloudWatch Logs

Need to process logs in real time

Trigger logic based on a specific error in logs

Evaluate Options:

Option A: SNS topic with filter policy

SNS filter policies work on message attributes, not on CloudWatch Logs subscription filters

Option B: Subscription filter on log group

This enables real-time log processing

You can create a subscription filter with a pattern matching specific error strings

Sends matched logs to a Lambda function or Kinesis

Option C: CloudWatch agent operator for trace collection

Irrelevant for log processing

Used for monitoring and tracing, not real-time log filtering

Option D: Lambda function to process logs

Once logs match the pattern, Lambda can process and act (e.g., alert, store, analyze)

Option E: EventBridge rule on a schedule

Not real-time

Scheduled EventBridge rules are for cron-like tasks, not log stream processing

Subscription filters: https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/SubscriptionFilters.html

Real-time log processing with Lambda: https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/SubscriptionFilters.html#LambdaExample

Logs in EKS to CloudWatch: https://docs.aws.amazon.com/eks/latest/userguide/fargate-logging.html

This solution will ensure that all orders and updates are stored to Amazon S3 with the least development effort because it uses DynamoDB Streams to capture changes in the DynamoDB table and trigger a Lambda function to write those changes to the S3 bucket. This way, the original Lambda function and API Gateway API endpoint do not need to be modified, and no additional services are required. Option A is not optimal because it will require more development effort to create a new Lambda function and a new API Gateway API endpoint, and to modify the original Lambda function to post updates to the new API endpoint. Option B is not optimal because it will introduce additional costs and complexity to use Amazon Kinesis Data Streams to create a new data stream, and to modify the Lambda function to publish orders to the data stream. Option D is not optimal because it will require more development effort to modify the Lambda function to publish to a new Amazon SNS topic, and to create and subscribe a new Lambda function to the topic.

Amazon S3 automatically scales to handle high request rates, but request parallelism is optimized when objects are distributed across multiple prefixes. AWS documentation explains that S3 internally partitions data by key name, and distributing objects across multiple prefixes allows requests to be spread across multiple partitions.

When many GET requests target objects within the same prefix, they may contend for the same internal partition, limiting throughput. By designing object keys with multiple prefixes—such as including hashed or randomized components—applications can significantly increase parallel request handling.

Options A and D do not address S3 request partitioning. Global Accelerator improves network latency for supported endpoints but does not increase S3 request parallelism. Direct Connect improves network consistency but does not change how S3 scales requests internally. Option B worsens the problem by concentrating traffic into a single prefix.

Therefore, partitioning objects by prefixes is the correct and AWS-recommended solution.

Amazon S3 supports resource-based policies, which are JSON documents that specify the permissions for accessing S3 resources. A resource-based policy can be used to enforce encryption in transit by denying access to requests that do not use HTTPS. The condition key aws:SecureTransport can be used to check if the request was sent using SSL. If the value of this key is false, the request is denied; otherwise, the request is allowed. Reference: How do I use an S3 bucket policy to require requests to use Secure Socket Layer (SSL)?

The solution that will meet the requirements is to use multiple stages in API Gateway. Create a Lambda function for each environment. Configure API Gateway stage variables to route traffic to the Lambda function in different environments. This way, the company can test the code in a dedicated, monitored test environment before releasing it to the production environment. The company can also use stage variables to specify the Lambda function version or alias for each stage, and avoid hard-coding the Lambda function name in the API Gateway integration. The other options either involve using a single stage in API Gateway, which does not allow testing in different environments, or adding different code blocks for different environments in the Lambda function, which increases complexity and maintenance.

For asynchronous Lambda invocations , AWS provides built-in failure handling options that require minimal code and minimal operational work. When an async invocation fails (after Lambda’s internal retry behavior), the event can be sent to a dead-letter queue (DLQ) so it is not lost. A DLQ is the standard mechanism for capturing events that could not be processed successfully, preserving the original payload for later inspection and reprocessing.

Using a DLQ (typically Amazon SQS or Amazon SNS ) gives durable storage of failed events and decouples failure recovery from the main execution path. The developer can later re-drive the messages by building a simple replay process, such as moving messages from the DLQ back to the original event source or invoking the function again with the stored payloads. This aligns with the requirement: “store messages that resulted in failed invocations so the application can retry later.”

Among the options, C is the only one that uses Lambda’s native async failure capture mechanism. Options A and B introduce unnecessary complexity and do not reliably store the original failed invocation payloads as a first-class workflow (CloudWatch Logs is not a message queue, and EventBridge→SNS doesn’t automatically capture only failed events from Lambda async invocations). Option D changes the architecture to an SQS pull model for the Lambda function; while valid, it is more than needed and adds design/operational considerations (polling, batching, visibility timeouts, DLQ configuration on the queue, etc.) compared with simply enabling DLQ support for async invokes.

Therefore, the least operational overhead solution is C : configure a dead-letter queue for the Lambda function’s asynchronous invocations so failed events are stored for later retry.

Amazon SNS supports server-side encryption (SSE) using AWS KMS to protect message content at rest . Enabling SSE on the SNS topic ensures that all messages published to the topic are encrypted using a customer-managed or AWS-managed KMS key. AWS documentation explicitly states that SSE is the mechanism for encrypting SNS messages stored within the service.

For encryption in transit , AWS services rely on HTTPS/TLS. To enforce this at the IAM policy level, AWS recommends using the global condition key aws:SecureTransport . By adding an explicit Deny statement when " aws:SecureTransport " : " false " , the policy ensures that any request not using HTTPS is rejected. This is a common AWS security best practice and is documented across multiple AWS services.

Option E correctly enforces encryption in transit. Option B is incorrect because denying when SecureTransport is true would block valid HTTPS requests. Option D is invalid because sns:Protocol applies to subscription protocols (such as email or HTTP endpoints), not to Lambda publishing actions. Option C (VPC endpoint) is not required because Lambda-to-SNS traffic is already encrypted over TLS by default.

Therefore, enabling SNS server-side encryption and enforcing HTTPS via an IAM deny condition fully satisfies the encryption-at-rest and encryption-in-transit requirements.

Requirement Summary:

Secrets managed in AWS Secrets Manager

DB: Amazon RDS for PostgreSQL

Need automated password rotation

Must maintain high availability

Least development effort

Rotation Strategies:

Single-user rotation strategy

Simplest to implement

The secret contains one set of credentials used by app and rotation logic

Supports automated rotation

AWS provides built-in Lambda rotation templates for RDS

A. Alternating-users strategy

⚠️ More complex

Requires application to switch users during rotation window

B. Manual secret + CLI rotation

Too much manual work

Not scalable or reliable

C. Multivalue answer rotation

Not a valid strategy in this context

Doesn’t apply to Secrets Manager

Secrets Manager rotation strategies: https://docs.aws.amazon.com/secretsmanager/latest/userguide/rotating-secrets.html

RDS PostgreSQL secret rotation: https://docs.aws.amazon.com/secretsmanager/latest/userguide/rotating-secrets_strategies.html#rotating-secrets-single-user

The hourly scheduled batch model is causing timeouts because the function must process an ever-growing backlog in a single invocation, which can exceed AWS Lambda’s maximum 15-minute runtime. A best-practice serverless refactor is to switch from “periodic polling” to an event-driven approach where work is split into smaller, independent units that can scale horizontally.

Configuring Amazon S3 Event Notifications to invoke Lambda on each object upload turns each file into its own discrete event. Instead of one long-running invocation that scans and processes many files, Lambda executes quickly per upload and automatically scales with incoming volume (subject to concurrency and downstream limits). This directly reduces processing time per run and eliminates the backlog buildup that leads to timeouts.

The remaining requirement is preventing duplicate processing. S3 event delivery is designed to be highly reliable, and applications should be built with idempotency in mind. Recording processed identifiers (such as a transaction ID, object key + version ID, or an application-level unique ID) in Amazon DynamoDB allows the function to check whether an item has already been processed before performing writes or downstream actions. Using DynamoDB with a conditional write (for example, “insert only if not exists”) is a common pattern to ensure exactly-once effect even when events are delivered more than once.

The other options do not meet the requirements as well: running more frequently (C) or continuously polling (D) still wastes time scanning and can still create duplicate work; moving to a single EC2 instance (B) reduces elasticity and increases operational overhead, and it doesn’t inherently solve deduplication or scaling.

Therefore, A provides scalable, event-driven processing and a straightforward deduplication mechanism.

Why Option B is Correct:Amazon ECS does not natively process docker-compose.yml files. Instead, the configurations from docker-compose.yml must be converted into ECS-compatible configurations within a task definition. Task definitions are the primary way to specify container configurations in ECS, including service dependencies like databases, caches, and volumes.

Steps to Resolve the Error:

Extract the configurations from the docker-compose.yml file.

Map the dependencies and settings to the appropriate ECS task definition fields.

Deploy the task definition to the ECS cluster.

Why Other Options are Incorrect:

Option A: Docker labels do not directly impact ECS task execution or integrate with ECS service configurations.

Option C: ECS namespaces do not exist as a feature.

Option D: Changing the service type to REPLICA does not resolve the issue of missing service dependency configurations.

AWS Documentation References:

Amazon ECS Task Definitions

Migrating Docker Compose Workloads to ECS

The HTTP header that the developer should use for this analysis is the X-Forwarded-For header. This header contains the IP address of the client that made the request to the Application Load Balancer. The developer can use this header to analyze patterns for the client IP addresses that use the application. The other headers either contain information about the protocol, host, or port of the request, which are not relevant for the analysis.

This solution allows the CloudFormation service role to access the S3 bucket from any account, as long as it has the S3 GetObject permission. The bucket policy grants access to any principal with the GetObject permission, which is the least privilege needed to deploy the Lambda code. This is more secure than granting ListBucket permission, which is not required for deploying Lambda code, or using a service-based link, which is not supported for Lambda functions.

To prove integrity (and authenticity) of data in transit, the correct cryptographic primitive is a digital signature. AWS KMS supports signing and verification operations using asymmetric KMS keys designed for signing (for example, RSA_SIGN_PSS, RSA_SIGN_PKCS1, or ECC_NIST_P256 key specs). The developer signs the data (or more commonly, a hash of the data) using the private key, and the external recipient verifies the signature using the corresponding public key. If the data is modified in transit, signature verification fails, proving the content was changed.

Option C exactly describes this model: sign with the private key, share the public key. The private key remains protected (ideally never leaving KMS). The public key can be distributed safely because it cannot be used to forge signatures.

Option A (encryption) provides confidentiality, not integrity proof to a third party, and sharing a symmetric encryption key is insecure and breaks key management principles.

Option B is incorrect because symmetric keys do not provide non-repudiation and generally require the verifier to possess the same secret key, which would allow the verifier to forge signatures too. While HMAC can validate integrity between trusted parties, it does not meet the typical “prove integrity” requirement to an external party without sharing a secret.

Option D is backward and insecure: you never share a private key.

Therefore, use an asymmetric KMS signing key to sign with the private key and provide the public key for verification.

Step-by-Step Breakdown:

Requirement Summary:

AWS AppSync API needs to invoke Lambda functions

Some Lambda functions are long-running

AppSync should return immediately, minimizing operational overhead

**Option A: AppSync + Lambda as data source using Event Mode

Correct: AWS AppSync supports asynchronous (event) invocation of Lambda data sources using Event Mode.

Event Mode means:

AppSync invokes Lambda asynchronously

Immediately returns a response to the client (typically a predefined payload or null)

Ideal for fire-and-forget workloads or when the response is not immediately needed

Option B: Increase Lambda timeout

Incorrect: This keeps AppSync waiting.

Even with increased timeout, synchronous invocations would block AppSync responses.

Option C: SQS queue + polling Lambda

Possible but too complex for this use case.

Requires additional infrastructure: queue + mapping + custom logic.

Higher operational overhead compared to built-in AppSync Event Mode.

Option D: Enable caching in AppSync

Irrelevant: AppSync cache is for optimizing repeated read queries, not for async workflows.

Asynchronous Lambda with AppSync: https://docs.aws.amazon.com/appsync/latest/devguide/resolver-mapping-template-reference-lambda.html#async-lambda-invocation

Lambda as AppSync Data Source: https://docs.aws.amazon.com/appsync/latest/devguide/tutorial-lambda-resolvers.html

Event Mode Docs: https://docs.aws.amazon.com/appsync/latest/devguide/lambda-resolvers.html#event-invocation-mode

AWS CloudFormation is a service that enables developers to model and provision AWS resources using templates. The developer can add a CloudFormation Deletion Policy attribute with the Retain value to the database resource. This will prevent the database from being deleted when the stack is deleted or updated. The developer can also update the CloudFormation stack policy to prevent updates to the database. This will prevent accidental changes to the database configuration or properties.

API Gateway’s canary deployments allow gradual traffic shifting to a new version of a function, minimizing disruption while testing.

Why Option A is Correct:

Gradual Rollout: Reduces risk by incrementally increasing traffic.

Rollback Support: Canary deployments make it easy to revert to the previous version.

Why Not Other Options:

Option B: Redeploying the stage disrupts all users.

Option C & D: Managing new stages and weighted routing introduces unnecessary complexity.

Canary Deployments in API Gateway

https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/access-control-overview.html#access-control-resource-ownership

To monitor end-to-end requests and debug issues across microservices, you need distributed tracing. Among AWS tools, AWS X-Ray is specifically built for this use case.

Step-by-Step Breakdown:

Step 1: Understand the Requirement

The developer needs:

End-to-end request tracing across microservices

Debugging capability for performance issues or failures

This points directly to a distributed tracing solution rather than just logs or metrics.

Step 2: Evaluate the Options

Option A: Amazon CloudWatch

CloudWatch is powerful for metrics, logs, and alerts.

But it does not provide distributed tracing or request path visualization between microservices.

So, it ' s not sufficient alone for the requirement.

Option B: AWS CloudTrail

CloudTrail tracks API calls made through the AWS Management Console, CLI, SDKs.

It is meant for auditing and governance, not microservices request tracing.

Not suitable for debugging microservices interactions.

Option C: AWS X-Ray SDK with X-Ray service map

Purpose-built for distributed tracing

Automatically collects data about requests as they travel through your microservices

You can instrument your code with the AWS X-Ray SDK in Java, Node.js, Python, Go, .NET, etc.

Visualizes requests using a service map, showing latency, errors, and time spent in downstream services.

How it works:

Instrument each microservice using the AWS X-Ray SDK.

Use the SDK to trace requests, propagate trace headers across services.

The X-Ray daemon collects and sends trace data to the X-Ray service.

You can view the service map, see bottlenecks, error rates, etc.

Option D: AWS Health

AWS Health provides information on AWS service outages or account-level events.

It does not monitor your application or microservices health or request flows.

Correct Choice: C is the only option that meets the developer ' s requirement to monitor end-to-end request paths and debug issues in a microservices architecture.

AWS Developer References:

AWS X-Ray Documentation: https://docs.aws.amazon.com/xray/latest/devguide/aws-xray.html

Instrumenting your application: https://docs.aws.amazon.com/xray/latest/devguide/xray-sdk.html

Viewing the Service Map: https://docs.aws.amazon.com/xray/latest/devguide/xray-console-service-map.html

Tracing Microservices with AWS X-Ray: https://aws.amazon.com/blogs/compute/tracing-microservices-aws-x-ray/

This step should be completed prior to deploying the application because it prepares the application artifacts for deployment. The AWS Serverless Application Model (AWS SAM) is a framework that simplifies building and deploying serverless applications on AWS. The AWS SAM CLI is a command-line tool that helps you create, test, and deploy serverless applications using AWS SAM templates. The sam package command bundles the application artifacts, such as Lambda function code and API definitions, and uploads them to an Amazon S3 bucket. The command also returns a CloudFormation template that is ready to be deployed with the sam deploy command. Compressing the application to a zip file and uploading it to AWS Lambda will not work because it does not use AWS SAM templates or CloudFormation. Testing the new Lambda function by first tracing it in AWS X-Ray will not prepare the application for deployment, but only monitor its performance and errors. Creating the application environment using the eb create my-env command will not work because it is a command for AWS Elastic Beanstalk, not AWS SAM.

Why Option C is Correct:

Amazon ElastiCache (Memcached) provides low-latency, in-memory caching suitable for session storage. It ensures stateless web tier operations and supports the high throughput of 5000 requests per minute.

Why Other Options are Incorrect:

Option A: RDS has higher latency compared to in-memory caching solutions like ElastiCache.

Option B: Shared file systems introduce additional complexity and are not ideal for low-latency session data storage.

Option D: DynamoDB has low latency but is less performant than ElastiCache for in-memory session management.

AWS Documentation References:

Amazon ElastiCache for Session State Management

Amazon Kinesis Data Firehose is a service that delivers real-time streaming data to destinations such as Amazon S3, Amazon Redshift, Amazon OpenSearch Service, and Amazon Kinesis Data Analytics. The developer can implement Kinesis Data Firehose data transformation as an AWS Lambda function. The function can remove pattern-based customer identifiers from the data and return the modified data to Kinesis Data Firehose. The developer can set an Amazon S3 bucket as the destination of the delivery stream.

AWS CodeBuild integrates natively with AWS Secrets Manager , allowing secrets to be securely injected into build environments without being exposed in logs. When referenced using the secrets-manager mapping in the env section of buildspec.yml, CodeBuild automatically retrieves and decrypts the secret at runtime.

AWS documentation states that Secrets Manager provides encryption at rest using AWS KMS, fine-grained IAM access control, and automatic rotation support. Secrets retrieved this way are masked in build logs by default, satisfying the requirement that secrets do not appear in logs.

Parameter Store (Option C) can store encrypted values, but Secrets Manager is AWS’s preferred service for managing secrets, especially when rotation and auditing are required. Options B and D introduce unnecessary steps and higher operational overhead.

Therefore, Secrets Manager with CodeBuild environment variables is the most secure and operationally efficient solution.

Amazon DynamoDB is a fully managed NoSQL database service that can store and retrieve any amount of data with high availability and performance. DynamoDB can handle concurrent requests from multiple IoT devices without throttling or data loss. To prevent duplicate requests from causing inconsistencies or data loss, the Lambda function can use DynamoDB conditional writes to check if the unique identifier for each request already exists in the table before processing the request. If the identifier exists, the function can skip or abort the request; otherwise, it can process the request and store the identifier in the table. Reference: Using conditional writes

The correct answer is C because for an in-place deployment on Amazon EC2 instances, AWS CodeDeploy automatic rollback works by redeploying the last known good revision when a deployment fails or when a configured monitoring alarm is triggered. AWS documentation explains that when automatic rollback is enabled, CodeDeploy creates a new deployment that uses the most recent successfully deployed application revision. This rollback deployment receives its own new deployment ID , because it is treated as a separate deployment event rather than a simple undo operation.

Option A is incorrect because CodeDeploy does not restore an Amazon S3 snapshot of the previous deployment state for EC2 in-place deployments. It redeploys the last successful revision instead. Option B describes behavior associated with blue/green deployment patterns , not in-place deployments . Route 53 alias switching and blue/green environment replacement are not how CodeDeploy handles failed validation during an EC2 in-place deployment. Option D is incorrect because CodePipeline orchestrates stages in a delivery pipeline, but it is not the service that performs the rollback behavior described here.

This distinction is important in AWS deployment design. In-place deployments update the same EC2 instances already serving the application. If validation fails, CodeDeploy does not roll traffic back through DNS or infrastructure replacement. Instead, it launches a rollback deployment of the previously successful application revision across those same instances. That behavior directly aligns with AWS CodeDeploy automatic rollback functionality for EC2/On-Premises in-place deployments.

Therefore, the correct result is that CodeDeploy redeploys the last known stable version as a new deployment , which makes C the right answer.

The application relies on shared file storage that behaves like NFS and must be accessible concurrently by multiple containers across Kubernetes nodes. AWS documentation clearly identifies Amazon EFS as the AWS-native, fully managed, NFS-compatible file system designed for this purpose.

Amazon EFS provides elastic scaling , high availability across multiple Availability Zones, and native integration with container orchestration platforms. It is the most direct replacement for on-premises NFS, requiring minimal architectural change.

To run Kubernetes workloads on AWS with high availability, AWS recommends Amazon EKS , a managed Kubernetes service that automatically handles control plane availability and upgrades. EKS integrates natively with EFS using the Amazon EFS CSI driver, allowing pods across multiple nodes to mount the same file system concurrently.

Amazon EBS (Options A, C, and D) is block storage that can be attached to only one EC2 instance at a time (except in limited Multi-Attach scenarios) and is therefore unsuitable for shared NFS-style access across a Kubernetes cluster.

Uploading container images to Amazon ECR ensures secure, scalable storage for container images and seamless integration with EKS.

Therefore, migrating the NFS data to Amazon EFS and running the workloads on Amazon EKS is the fastest, most scalable, and AWS-recommended solution.

The goal is near-real-time propagation of changes from one DynamoDB table (Accounts) to another service’s datastore (Payments) while keeping services decoupled. The simplest and most reliable DynamoDB-native change feed is Amazon DynamoDB Streams.

DynamoDB Streams captures item-level modifications (inserts, updates, deletes) in time order and makes them available as a stream of change records. A consumer (commonly an AWS Lambda function) can process stream records and apply the necessary updates to the Payments database/table (or publish events for downstream processing). This provides near-real-time updates with minimal operational overhead, strong integration, and a decoupled architecture because the Accounts service simply emits changes and does not need to call Payments synchronously.

Option A (Glue ETL) is batch-oriented and adds heavy operational complexity; it’s not the simplest nor near-real-time.

Option B introduces a cache as a synchronization mechanism, which complicates consistency and durability; caches are not ideal as the primary propagation channel.

Option C (Firehose) is designed for delivering streaming data to destinations like S3, Redshift, OpenSearch, etc., and is not the standard approach for DynamoDB-to-DynamoDB change replication.

Therefore, DynamoDB Streams is the best fit for near-real-time, decoupled updates between DynamoDB-backed microservices.

CloudFormation and Lambda Environment Variables:

CloudFormation is an excellent tool to manage infrastructure as code, including the log group resource.

Lambda functions can access environment variables at runtime, making them a suitable way to pass configuration information like the log group ARN.

CloudFormation Template Modification:

In your CloudFormation template, define the log group resource.

In the Lambda function resource, add an Environment section:

YAML

Environment:

Variables:

LOG_GROUP_ARN: !Ref LogGroupResourceName

Use code with caution.

content_copy

The !Ref intrinsic function retrieves the log group ' s ARN, which CloudFormation generates during stack creation.

Using the ARN in Your Lambda Function:

Within your Lambda code, access the LOG_GROUP_ARN environment variable.

Configure your logging library (e.g., Python ' s logging module) to send logs to the specified log group.

The correct answer is A. Add a configuration file in TOML format to group configuration entries to every environment. Add a table for each testing and staging environment. Deploy updates to the environments by using the sam deploy command and the --config-env flag that corresponds to the each environment.

A. Add a configuration file in TOML format to group configuration entries to every environment. Add a table for each testing and staging environment. Deploy updates to the environments by using the sam deploy command and the --config-env flag that corresponds to the each environment. This is correct. This solution will meet the requirements with the least development effort, because it uses a feature of the AWS SAM CLI that supports a project-level configuration file that can be used to configure AWS SAM CLI command parameter values1. The configuration file can have multiple environments, each with its own set of parameter values, such as stack name, region, capabilities, and more2. The developer can use the --config-env option to specify which environment to use when deploying the application3. This way, the developer can avoid creating multiple templates or scripts, or manually overriding parameters for each environment.

B. Create additional AWS SAM templates for each testing and staging environment. Write a custom shell script that uses the sam deploy command and the --template-file flag to deploy updates to the environments. This is incorrect. This solution will not meet the requirements with the least development effort, because it requires creating and maintaining multiple templates and scripts for each environment. This can introduce duplication, inconsistency, and complexity in the deployment process.

C. Create one AWS SAM configuration file that has default parameters. Perform updates to the testing and staging environments by using the —parameter-overrides flag in the AWS SAM CLI and the parameters that the updates will override. This is incorrect. This solution will not meet the requirements with the least development effort, because it requires manually specifying and overriding parameters for each environment every time the developer deploys the application. This can be error-prone, tedious, and inefficient.

D. Use the existing AWS SAM template. Add additional parameters to configure specific attributes for the serverless function and database table resources that are in each environment. Deploy updates to the testing and staging environments by using the sam deploy command. This is incorrect. This solution will not meet the requirements with the least development effort, because it requires modifying the existing template and adding complexity to the resource definitions for each environment. This can also make it difficult to manage and track changes across different environments.

Amazon EventBridge: A service that reacts to events from various AWS sources, including S3. Rules define which events trigger actions (like invoking Lambda functions).

S3 Object Created Events: EventBridge can detect these, providing seamless integration for automated CSV processing.

S3 Lifecycle Rules: Allow for actions based on object age or prefixes. These can directly trigger Lambda functions for file processing.

The solution that will meet the requirements is to create an Amazon EventBridge rule that runs on a regular schedule to invoke the Lambda function. This way, the developer can use an automated and serverless way to invoke the function every 10 minutes. The developer can also use a cron expression or a rate expression to specify the schedule for the rule. The other options either involve using an Amazon EC2 instance, which is not serverless, or using environment variables or query parameters, which do not trigger the function.

Step-by-Step Breakdown:

Requirement Summary:

Get notified when EC2 CPU Utilization > 80%

Option A: CloudWatch Alarm with SNS

Correct and standard AWS practice

CloudWatch automatically collects EC2 metrics, including CPUUtilization.

You can set a CloudWatch Alarm with a threshold (80% in this case).

Then, trigger an SNS notification to email, SMS, Lambda, etc.

Option B: AWS CloudTrail alarm

Incorrect: CloudTrail logs API activity, not performance metrics.

It doesn’t track metrics like CPU utilization.

Option C: Cron job on EC2 running --describe-instance-information

Incorrect: This doesn’t give CPU usage.

Also inefficient, and polling is bad practice when CloudWatch already monitors this natively.

Option D: Lambda function querying CloudTrail for CPU usage

Incorrect and conceptually flawed.

CloudTrail does not store performance metrics; CloudWatch does.

CloudWatch Alarms for EC2: https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/AlarmThatSendsEmail.html

EC2 Metrics in CloudWatch: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/viewing_metrics_with_cloudwatch.html

Amazon SNS Notification Setup: https://docs.aws.amazon.com/sns/latest/dg/sns-email-notifications.html

CodeDeploy Predefined Configurations: CodeDeploy offers built-in deployment configurations for common scenarios.

Canary Deployment: Canary deployments gradually shift traffic to a new version, ideal for controlled rollouts like this requirement.

CodeDeployDefault.ECSCanary10Percent15Minutes: This configuration matches the company ' s requirements, shifting 10% of traffic initially and then completing the rollout after 15 minutes.

Mocking API Responses: API Gateway ' s Mock integration type enables simulating API behavior without invoking backend services.

Testing with Sample Data: Using captured responses from the real third-party API ensures realistic testing of the integration code.

Focus on Integration Logic: This solution allows the developer to isolate and test the application ' s interaction with the payment APIs, even without a test environment from the third-party providers.

https://aws.amazon.com/blogs/devops/automated-code-review-on-pull-requests-using-aws-codecommit-and-aws-codebuild/

Amazon CloudWatch is a service that monitors AWS resources and applications. The developer can use CloudWatch to monitor and troubleshoot all applications from one place. To do so, the developer needs to download the CloudWatch agent to the on-premises server and configure the agent to use IAM user credentials with permissions for CloudWatch. The agent will collect logs and metrics from the on-premises server and send them to CloudWatch.

AWS SAM provides built-in tooling to generate realistic event payloads that match the structure of events produced by AWS services (such as S3, SNS, SQS, API Gateway, EventBridge, and others). The command designed specifically for this purpose is sam local generate-event. It produces sample event JSON that mirrors the format AWS services deliver to Lambda, which meets the requirement that payload structures match real service events.

This approach also requires the least development effort because the developer does not need to manually craft and maintain JSON payload files. Instead, the developer can generate the appropriate event type with SAM, optionally tweak values (bucket name, object key, request parameters), and then run sam local invoke using the generated payload. This significantly reduces mistakes and saves time, especially when testing multiple event sources.

Why other options are less suitable:

A (shareable test events) is vague and typically still requires manual creation/maintenance.

B requires manually creating payloads, which is error-prone and higher effort.

C adds unnecessary operational overhead (S3 storage) and still relies on manual payload creation.

Therefore, using sam local generate-event is the quickest and lowest-effort way to generate accurate AWS-service-like test payloads for local Lambda testing.

Using Amazon ElastiCache for Memcached to store and manage the session data will reduce the memory load and improve the performance of the web server. Using Amazon RDS for MySQL DB instance to store the application data will provide a scalable, reliable, and managed database service. Option A is not optimal because it does not address the memory issue of the web server. Option C is not optimal because it does not provide a persistent storage for the application data. Option D is not optimal because it does not provide a high availability and durability for the session data.

AWS Lambda automatically integrates with Amazon CloudWatch Logs, but logging only works if the function’s execution role has the correct IAM permissions . AWS documentation states that Lambda functions require permissions for logs:CreateLogGroup, logs:CreateLogStream, and logs:PutLogEvents to successfully publish log entries.

In this scenario, CloudWatch metrics show errors, which confirms that the function is executing. The absence of logs despite logging statements in the code strongly indicates a missing IAM permission issue , not a runtime or tracing issue.

CloudWatch Lambda Insights (Option B) provides performance metrics but does not enable basic logging. AWS X-Ray (Option C) provides distributed tracing but does not replace CloudWatch Logs. Option D is invalid because CloudWatch does not assume roles to collect logs.

AWS documentation explicitly identifies missing log permissions as the most common cause of absent Lambda logs. Therefore, granting the Lambda execution role the appropriate CloudWatch Logs permissions resolves the issue.

The AWS Serverless Application Model Command Line Interface (AWS SAM CLI) is a command-line tool for local development and testing of Serverless applications2. The sam local start-api subcommand of AWS SAM CLI is used to simulate a REST API by starting a new local endpoint3. Therefore, option D is correct.

The correct answer is A because the requirement is to encrypt data in transit and data at rest while ensuring that the developer can manage encryption keys . AWS Key Management Service (AWS KMS) is the AWS-managed service specifically designed for creating, controlling, and auditing cryptographic keys. When used with the AWS Encryption SDK , it supports envelope encryption , a recommended pattern for protecting sensitive application data.

In this model, the application uses data keys to encrypt the actual sensitive data, while AWS KMS protects and manages the master keys that encrypt those data keys. This provides strong security and centralized key management, including access control through IAM, key rotation options, and auditability through AWS logging integrations. For data in transit , the application should also use TLS/HTTPS , which is standard AWS guidance for secure communications.

Option B is incorrect because Parameter Store is not the correct mechanism for directly managing application encryption keys in this way. SecureString protects stored values, but it is not a substitute for using KMS as the primary managed key service for encryption architecture. Option C is incorrect because Secrets Manager is intended for storing secrets such as passwords and tokens, not as the main key management system for application encryption design. Option D is incorrect because SSE-S3 uses Amazon S3-managed keys, not customer-managed KMS keys, so it does not satisfy the requirement that the developer be able to manage encryption keys.

Therefore, AWS KMS with envelope encryption through the AWS Encryption SDK is the most appropriate and secure solution.

The requirement is for a single secret value to be available in multiple AWS Regions while remaining consistent across those Regions, with the least operational overhead . AWS Secrets Manager provides a managed capability for this use case: secret replication (also referred to as multi-Region secrets). When replication is enabled from a primary Region, Secrets Manager automatically creates and maintains synchronized replicas of the secret in the selected secondary Regions. Updates to the primary secret are propagated to the replicas so the secret value remains consistent across Regions.

This approach (option C) minimizes operational effort because AWS manages the replication workflow, consistency, and the lifecycle of replicas. The application in each Region can retrieve the secret from the local Region endpoint of Secrets Manager by referencing the Region-appropriate secret ARN, improving latency and reducing cross-Region dependency. It also improves resilience: if a Region has issues, workloads in another Region can still read the replicated secret locally.

Option A recreates managed replication with custom glue (Lambda + EventBridge), which adds code, scheduling, error handling, permissions, retries, and monitoring—higher operational overhead and more failure modes. Option B introduces a cross-Region runtime dependency on the primary Region and increases latency; local caching also risks stale secret values during rotations/updates unless carefully managed, and it reduces resilience if the primary Region is impaired. Option D (independent secrets per Region) increases operational overhead and directly risks inconsistency, because keeping values synchronized becomes a manual or custom automated process.

Therefore, C is the best solution: enable Secrets Manager replication in the primary Region and have applications in each Region read from their local replicated secret ARN to achieve consistent, multi-Region availability with minimal operational work.

The solution that will meet the requirements with the most operational efficiency is to create a new CodePipeline stage that occurs after the container image is built. Configure ECR basic image scanning to scan on image push. Use an AWS Lambda function as the action provider. Configure the Lambda function to check the scan results and to fail the pipeline if there are findings. This way, the container image is analyzed earlier in the CI/CD pipeline and any vulnerabilities are detected and reported before deploying to the EKS cluster. The other options either delay the analysis until after deployment, which increases the risk of exposing insecure images, or perform analysis on the source code instead of the container image, which may not capture all the dependencies and configurations that affect the security posture of the image.

IAM Roles for EC2: IAM roles are the recommended way to provide AWS credentials to applications running on EC2 instances. Here ' s how this works:

You create an IAM role with the necessary permissions to access the target S3 bucket.

You create an instance profile and associate the IAM role with this profile.

When launching the EC2 instance, you attach this instance profile.

Temporary Security Credentials: When the application on the EC2 instance needs to access S3, it doesn ' t directly use access keys. Instead, the AWS SDK running on the instance retrieves temporary security credentials associated with the role. These are rotated automatically by AWS.

The requirement is encryption in transit for sensitive data sent in API calls between two Apache-based EC2 fleets in peered VPCs . The most operationally efficient approach is to restrict network access and rely on standard TLS at the application layer. Among the provided options, the only operationally simple and directly relevant network control is security group referencing across VPC peering within the same account and Region.

By creating security groups that allow inbound traffic only from the other fleet’s security group , the developer ensures that only the intended instances can communicate over the required ports (for example, 443 for HTTPS). This is operationally efficient because it avoids managing IP allowlists that change with Auto Scaling and keeps the trust boundary tied to instance membership rather than addresses. It also supports least privilege by limiting which sources can reach the service.

Why the other options are not the best fit:

B (Site-to-Site VPN) is unnecessary for VPC peering in the same account/Region and adds significant operational overhead. It also does not replace the need for TLS at the application layer to guarantee encryption to the endpoint.

C (EBS encryption) addresses encryption at rest on volumes, not encryption in transit between fleets.

D (Nitro Enclaves) is heavy and unrelated to the core requirement; enclaves protect data-in-use and do not replace standard TLS for network encryption.

Important nuance: Security groups alone do not encrypt traffic; encryption is achieved by using HTTPS/TLS between the Apache services. However, option A is the most operationally efficient control offered here to implement secure connectivity between the fleets (paired with using HTTPS on Apache, which is the standard for encrypting API calls).

Therefore, A best meets the requirement with the least operational complexity.

Browsers enforce the Same-Origin Policy, which restricts web pages from requesting resources (like JavaScript files, fonts, and other assets) from a different domain/origin unless the target explicitly allows it. When the developer moved shared JavaScript and web fonts into a separate “central” S3 bucket, the applications likely began fetching these assets from a different origin than the app’s primary origin (for example, different domain, subdomain, or CloudFront distribution). As a result, the browser blocks those requests unless the response includes the appropriate CORS headers.

AWS documentation for Amazon S3 explains that to allow cross-origin requests from browsers, you must configure Cross-Origin Resource Sharing (CORS) on the S3 bucket that serves the resources. A CORS configuration defines allowed origins, methods (such as GET), and headers. For fonts in particular, browsers are strict and commonly require CORS headers to load fonts across origins.

Option C directly addresses this by adding a CORS policy to the central bucket, allowing the subsidiary application origins to retrieve JS and font files without being blocked.

Option B (bucket policy) controls authorization at the AWS request level, but it does not instruct browsers to allow cross-origin access. Even if the request is authorized, the browser can still block it if CORS headers are missing.

Option A (access points) is also about access management and does not solve browser CORS enforcement.

Option D (Content-MD5) is for integrity checking and is unrelated to cross-origin browser blocking.

Lambda throttling occurs when the number of concurrent executions exceeds the available concurrency. This can happen due to account-level concurrency limits, function-level reserved concurrency limits, or sudden traffic spikes.

The most operationally efficient ways to reduce throttling are to increase available concurrency:

Option C: Increasing the function’s reserved concurrency can reduce throttling when the function is being constrained by a too-low reserved concurrency value. Reserved concurrency guarantees a fixed amount of concurrency for the function (and also caps it). If the cap is currently too low, raising it allows more parallel executions and reduces throttles.

Option E: If throttling is due to the account concurrency quota being reached (or burst scaling limits in some patterns), the correct fix is to request a service quota increase. This increases the total available concurrency capacity for the account (and/or specific dimensions), allowing more simultaneous executions across functions.

Why the others are not correct:

A (migrate to EKS) is high effort and not operationally efficient for solving Lambda throttling.

B (maximum age of events) applies to asynchronous event retries/queues; it does not reduce throttling—it only changes how long events remain eligible for processing.

D is irrelevant: adding lambda:GetFunctionConcurrency to the execution role only allows reading settings and does not change throttling behavior.

Therefore, the best operational fixes are to increase reserved concurrency and request a concurrency quota increase.

Amazon S3 Lifecycle rules are specifically designed to manage object versions over time. When versioning is enabled, lifecycle rules can be configured to expire noncurrent versions , retaining only a defined number of recent versions.

AWS documentation states that lifecycle rules can be used to “retain a specific number of newer noncurrent versions.” By configuring a lifecycle rule to keep one noncurrent version , Amazon S3 automatically deletes older versions while retaining the current and immediately previous versions.

Bucket policies (Option A) control access, not retention. S3 Object Lock (Option C) is designed for regulatory compliance and write-once-read-many (WORM) use cases and is not suitable for routine version cleanup. Suspending versioning (Option D) does not remove existing versions and increases application complexity.

Therefore, an S3 Lifecycle rule is the correct and AWS-recommended solution.

This solution will meet the requirements with the least operational overhead because it uses Amazon S3 as a scalable, secure, and durable storage service for the reports. The presigned URL will allow customers to access their reports for a limited time (8 hours) without requiring additional authentication. The S3 Lifecycle configuration rules will automatically delete the reports that are older than 2 days, reducing storage costs and complying with the data retention policy. Option A is not optimal because it will incur additional costs and complexity to store the reports as DynamoDB items, which have a size limit of 400 KB. Option B is not optimal because it will not provide customers with access to their reports within one hour, as Amazon SNS email delivery is not guaranteed. Option D is not optimal because it will require more operational overhead to manage an RDS database and a Lambda function for storing and deleting the reports.

To measure how long an application spends calling downstream AWS services (for example, DynamoDB, S3, SNS, etc.) in a distributed system, the most effective AWS-native approach is AWS X-Ray with appropriate instrumentation. X-Ray provides distributed tracing by capturing end-to-end request paths and breaking each request into segments and subsegments with precise timing information. When the application is instrumented with the X-Ray SDK, client handlers/interceptors automatically create subsegments for calls to AWS services, recording latency, response status, and error/throttle indicators.

This directly satisfies the requirement: determine the length of time of calls to various downstream services. In the X-Ray trace map and trace details, the developer can see per-service timing, identify which dependency is slow, and distinguish application processing time from downstream call time. This is specifically designed for performance bottleneck analysis and root cause investigation across distributed components.

Option A (VPC Flow Logs) captures network flow metadata (source/destination, ports, bytes, accept/reject), not application-level request timing to AWS service APIs, and it won’t show the latency breakdown per downstream service call.

Option B can work only if the application is already logging detailed timing for each call, but this is manual, inconsistent across services, and does not give a unified distributed trace view. It is also higher effort than using standard tracing instrumentation.

Option C increases CloudWatch metric granularity for EC2 but does not show per-request downstream service call durations.

Therefore, implementing AWS X-Ray with client handlers is the correct solution.

The correct answer is A because AWS CodeBuild natively integrates with AWS Secrets Manager for securely injecting secrets into build environments. In the env section of buildspec.yml , CodeBuild supports a secrets-manager mapping that references a secret directly. This approach keeps the secret encrypted at rest in Secrets Manager, makes it available to the build only when needed, and minimizes custom retrieval logic and operational complexity.

This solution also aligns with the requirement that the secret must not appear in the build logs. AWS CodeBuild is designed to handle environment variables sourced from supported secret stores in a way that avoids exposing plaintext secret values during normal build output. By granting the minimum necessary IAM permissions to the CodeBuild service role, access to the secret is tightly controlled according to least-privilege principles.

Option B is not the best answer because storing secrets in S3 and downloading them during the build adds unnecessary operational overhead and increases the chance of accidental exposure. NoEcho is associated with CloudFormation parameters, not with securely masking S3-downloaded secrets in CodeBuild logs. Option C can also work because Parameter Store integrates with CodeBuild, but for secrets specifically, Secrets Manager is the more direct AWS-managed secret storage service and is generally preferred for this use case. Option D is less efficient because custom CLI retrieval in pre-build steps adds more code and more operational work than native integration.

Therefore, the solution with the least operational overhead and strong security is to use Secrets Manager with the buildspec.yml secrets-manager mapping , making A the correct answer.

The requirements describe a real-time streaming use case where multiple processing applications must read the same stream in parallel, each must be able to resume processing after interruptions without data loss, and the system should support adding new processors with minimal duplication. Amazon Kinesis Data Streams (KDS) is purpose-built for this pattern.

With Kinesis Data Streams, producers write records to a stream that is split into shards. Multiple consumer applications can independently read from the stream using separate consumer groups (for example, using Kinesis Client Library). Each consumer maintains its own checkpoint (sequence number position), which allows it to resume from the last processed record after a restart or failure. This directly meets the “resume without losing data” requirement.

KDS also supports a configurable retention period (commonly 24 hours by default, extendable), enabling replay and recovery without forcing producers to re-send data or duplicating messages for each processor. Adding new consumers is straightforward: a new processor can begin reading the existing stream (from latest or from a trim horizon), which minimizes duplication compared to fan-out duplication patterns.

Why the others don’t fit:

SQS is a queue, not a stream. Multiple consumers typically compete for messages rather than all consuming the same messages independently, and duplication is required if multiple apps must process the same data.

Kinesis Data Firehose is primarily for delivery to destinations (S3, Redshift, OpenSearch) and not designed for multiple real-time parallel consumers with replay semantics.

EventBridge is event routing with limited replay/resume semantics compared to KDS and is not intended for high-throughput stream processing with per-consumer checkpoints.

Therefore, Amazon Kinesis Data Streams satisfies all requirements.

AWS Amplify’s most direct support for GraphQL APIs is through AWS AppSync, and the Amplify CLI can generate and configure an AppSync GraphQL API directly from a schema with minimal setup. The requirement says the API already has an existing GraphQL schema, and the goal is to migrate it with the least configuration effort.

Option D is the simplest: run amplify add api, choose GraphQL, and provide the existing schema. Amplify then provisions the AppSync API, sets up the schema, creates the backend resources (depending on chosen data sources), and wires the configuration into the Amplify project so the SPA can consume the API.

Option A is incorrect because it selects REST and does not align with an existing GraphQL schema.

Option B is incorrect because API Gateway is not the native GraphQL service and would require additional mapping/proxy logic—more configuration.

Option C can be valid if an AppSync API already exists and you want to import it, but the question asks to “migrate the API along with the application” with least configuration. Creating it directly in Amplify is typically less configuration than creating separately and importing.

Therefore, using Amplify CLI to add a GraphQL API and supply the existing schema is the least-config approach.

The base table’s primary key is UserID, which means efficient Query operations on the table natively require specifying UserID (partition key equality). The required access pattern, however, is to retrieve many users for a given Location and then filter/sort by RegistrationDate (users who registered after a given date). Because Location is not part of the table’s primary key, the efficient DynamoDB approach is to create an index that supports this access pattern.

A global secondary index (GSI) can have a different partition key and sort key than the base table. By creating a GSI with Location as the partition key and RegistrationDate as the sort key, the application can use the Query operation to fetch all users in a specific location (Location = :loc) and apply a sort key condition such as RegistrationDate > :date (or BETWEEN), which is efficient and avoids a full-table scan. This design is the canonical pattern for optimizing DynamoDB queries: model the table and indexes around known access patterns so the application can use Query instead of Scan.

Option C (LSI) is not valid for this requirement because an LSI must use the same partition key as the base table (UserID). Since the access pattern is by Location, an LSI cannot make Location the partition key. Option B (Scan + filter) is inefficient at scale because Scan reads every item (or every item in the scanned segments) and then filters results, consuming read capacity and increasing latency. Option D (BatchGetItem) requires the caller to already know the primary keys of the items to retrieve; it cannot discover “all users in a location after a date” and does not support filtering in the way Query on an index does.

Therefore, A is the correct and most efficient solution: a GSI on (Location, RegistrationDate) and Query with a date range condition.

Why Option A is Correct:Converting a Lambda function to use a container image involves packaging the function code into a container image, storing the image in Amazon Elastic Container Registry (ECR), and updating the function to use the ECR repository URI.

Why Other Options are Incorrect:

Option B: SAM templates support container-based Lambda deployment, but storing the image in S3 is not applicable.

Option C: CloudFormation does not natively support specifying Lambda container images in S3.

Option D: While partially correct, it omits the need to specify the image tag for the deployment.

AWS Documentation References:

Lambda Container Images

An Application Load Balancer can invoke Lambda functions as targets, but ALB must be explicitly granted permission to invoke the function. This is done by adding a resource-based permission to the Lambda function policy that allows the Elastic Load Balancing service principal to call lambda:InvokeFunction, typically scoped to the specific target group ARN.

If the developer registers the function as a target but does not add the required permission, the ALB will not be able to invoke the function when requests arrive. This is a common integration oversight: the target registration succeeds, but invocation fails because Lambda denies the call.

Option A is incorrect because ALB does support Lambda targets.

Option B is incorrect because Lambda targets can be configured through the console, CLI, or APIs.

Option D is unrelated: cross-zone load balancing affects distribution of traffic across AZs for instance/IP targets, not whether Lambda invocation is authorized.

Amazon S3 is a service that provides highly scalable, durable, and secure object storage. The developer can create an S3 bucket to host the repository and migrate the existing .xml files to the S3 bucket. The developer can update the application code to use the AWS SDK to read and write configuration files from S3. This solution will meet the requirement of high availability for the repository in a cost-effective way.

The correct answer is D because Amazon CloudFront is the AWS content delivery network (CDN) designed to improve performance for users who access content globally. In this scenario, the application serves millions of video clips from Amazon S3 to users around the world, and the main problem is poor streaming quality as traffic scales. CloudFront addresses this by caching content at edge locations that are geographically closer to users, reducing latency and improving throughput.

AWS documentation emphasizes that CloudFront works especially well with Amazon S3 as an origin for static and streaming content. Instead of every viewer retrieving video data directly from the S3 bucket’s Region, CloudFront serves cached objects from nearby edge locations. This provides a major performance boost for global audiences and helps absorb large traffic spikes more efficiently.

Option A is incorrect because Route 53 geolocation routing directs DNS queries based on location, but it does not cache or accelerate video delivery. Option B can improve regional redundancy, but simply replicating objects across Regions does not automatically provide fast global delivery or edge caching. Option C is about storage cost optimization, not delivery performance.

CloudFront is also well suited for sudden spikes in traffic because AWS’s edge network is built for high-scale distribution. For video streaming workloads, the reduction in latency and the ability to serve content from nearby cached locations usually provide the largest performance improvement compared to the other options.

Therefore, the best solution is to create an Amazon CloudFront distribution with Amazon S3 as the origin , making D the correct answer.

The correct answers are D and E .

The requirement states that the company does not want to manage the underlying hosts . That means the containerized application should run on AWS Fargate , which is the serverless compute engine for Amazon ECS. With Fargate, AWS manages the underlying infrastructure, so the company does not need to provision, patch, or scale EC2 instances. To run a task on Fargate, the task definition must specify FARGATE in the requiresCompatibilities field. That is why D is correct.

The developer also needs to monitor CPU and memory utilization in the Amazon CloudWatch console . For ECS tasks on Fargate, the task definition must specify valid CPU and memory values. These values define the task’s resource allocation and enable ECS and CloudWatch to report utilization metrics against those configured resources. Therefore, E is also required.

Option A applies to ECS on EC2, where the company would manage cluster instances, which violates the requirement. Option B is unnecessary because ECS and CloudWatch already provide the needed service metrics without requiring a custom script. Option C is incorrect because the task does not need cloudwatch:GetMetricData permission merely for CloudWatch to display CPU and memory utilization metrics.

This design follows AWS best practices for running containers without host management while still gaining built-in observability. By selecting Fargate and defining the CPU and memory values in the task definition, the application can run serverlessly on ECS and expose the required utilization metrics in CloudWatch.

Therefore, the correct combination is D and E .

The solution that will meet the requirements is to add a test phase to the amplify.yml build settings for the application. This way, the developer can run end-to-end tests on every code commit and catch any bugs before deploying to production. The other options either do not support end-to-end testing, or do not run tests automatically.

To determine who deleted an Amazon RDS DB instance, the correct source of truth is AWS CloudTrail, which records API activity and includes the identity (IAM user, role, assumed role session) that made the call. Deleting an RDS instance is performed through the RDS API action DeleteDBInstance, and CloudTrail logs an event for this action that contains key fields such as userIdentity, eventTime, eventName, sourceIPAddress, and request parameters identifying the DB instance (for example, dBInstanceIdentifier).

Because the DB instance was deleted within the past 90 days, CloudTrail Event History is commonly sufficient (Event History typically retains 90 days of management events). If the company has a CloudTrail trail configured to deliver logs to S3/CloudWatch Logs, those logs can also be queried for the same event.

Option A directly matches this: retrieve CloudTrail events for DeleteDBInstance related to mysql-db and inspect the userIdentity section to identify the IAM principal that performed the deletion.

Option B is not reliable because RDS log groups (when enabled) capture database engine logs (slow query log, error log, general log) and do not record control-plane actions like who deleted the instance.

Option C is incorrect because X-Ray traces application-level request flows; it does not audit administrative actions like RDS deletion.

Option D is not applicable: Systems Manager inventory does not provide authoritative records of RDS deletions or the IAM principal responsible.

Therefore, CloudTrail lookup for DeleteDBInstance events is the correct solution.

Comprehensive and Detailed Step-by-Step Explanation:

To confirm that the HTTP headers, body, and responses meet expectations, you need to test the specific HTTP Task state in isolation and inspect the details.

Option A: TestState API with TRACE:

The TestState API allows developers to test individual states in a state machine without executing the entire workflow.

Setting the inspection level to TRACE provides detailed information about the HTTP request and response, including headers, body, and status codes.

This option provides the precise and granular testing required to verify the HTTP Task functionality.

Why Other Options Are Incorrect:

Option B: The DEBUG inspection level provides less detailed information than TRACE and focuses on general debugging, not a detailed view of HTTP interactions.

Option C: Step Functions does not have a " data flow simulator " to test individual tasks; this option is not valid.

Option D: Changing the state machine’s log level to ALL increases logging granularity for the entire state machine but does not allow isolated testing of a specific HTTP Task.

Lambda Concurrency: The ' maximum concurrency ' setting in event source mappings controls the maximum number of simultaneous invocations Lambda allows for that specific source.

Prioritizing Queues: Setting a lower maximum concurrency for the ' high priority queue ' ensures it ' s processed first while allowing more concurrent invocations from the ' low priority queue ' .

Batching: Batch size settings affect the number of messages Lambda retrieves from a queue per invocation, which is less relevant to the prioritization requirement.

Requirement Summary:

Orders are being processed more than once

Need to prevent duplicate processing

Looking for least implementation effort

Key Concept:

Lambda + Event-driven patterns can occasionally result in duplicate invocations (at-least-once delivery model)

You need idempotency (i.e., prevent repeated processing of same event)

Evaluate Options:

A. Use DynamoDB for de-duplication

Simple and widely used approach

Store a unique orderId as the primary key

Before processing, check if order exists

If yes → skip

If no → process and store the ID

Minimal code changes required

B. ECS + Step Functions

Overkill for basic de-duplication

Adds significant complexity

C. Retry logic with fixed delay

Doesn ' t prevent duplication — makes it worse

Retrying might trigger the same message again

D. Athena to identify duplicates

Reactive solution, not preventative

Not suitable for real-time event de-duplication

Lambda idempotency: https://docs.aws.amazon.com/lambda/latest/dg/invocation-retries.html

Using DynamoDB for idempotent design: https://aws.amazon.com/blogs/compute/how-to-design-idempotent-APIs-on-aws/

Why Option B is Correct:The RetainExceptOnCreate deletion policy ensures that the IAM user is retained after successful stack creation but is deleted if the stack creation fails or rolls back. This meets both requirements.

Why Other Options are Incorrect:

Option A: The Retain policy retains the resource regardless of stack status and does not delete the IAM user upon rollback.

Option C: Updating the service role for termination protection does not address the specific deletion behavior for the IAM user.

Option D: Stack policy controls updates, not resource deletion behavior during rollbacks.

AWS Documentation References:

CloudFormation DeletionPolicy Attribute