Free DVA-C02 flashcards

Which Amazon SQS configuration gives ordered, exactly-once processing per customer account so retried sends never double-charge?

Use an SQS FIFO queue with the customer account as the message group ID and content-based deduplication enabled. FIFO preserves order within each message group, and deduplication discards repeated sends inside a five-minute window, so each account's transactions process in order and exactly once.

How do you fan out a copy of every event to several independent consumers, including SQS queues and an HTTP endpoint, on AWS?

Publish to an Amazon SNS topic and subscribe each consumer to it. SNS pushes a separate copy of every message to each subscriber, so subscribed SQS queues buffer events for their own pollers and an HTTP subscriber receives push delivery with retries. Each consumer then processes at its own pace.

Common misconceptionThat a shared SQS queue gives every consumer a copy of each message. A standard queue delivers each message once, so pollers compete rather than each receiving it.

How does a Cognito-authenticated device obtain temporary AWS credentials to call Amazon S3 directly without embedding access keys?

Configure an Amazon Cognito identity pool that trusts the user pool as an authentication provider. It exchanges the validated user pool token for temporary credentials from an assumed IAM role via AWS STS. The AWS SDK signs S3 calls with those rotating credentials, so no long-lived key is stored.

Common misconceptionThat an IAM access key shipped inside a distributed mobile app is an acceptable way to authorise uploads. It can be extracted from the bundle and never rotates.

Which Amazon Cognito component handles end-user sign-up, sign-in, and password reset without a custom user database?

An Amazon Cognito user pool is a fully managed identity directory that runs sign-up, sign-in, and password recovery flows. On successful authentication it issues JWT ID and access tokens. This removes the need to build and operate your own user store for credential management.

Common misconceptionThat every application end user should get their own IAM user for sign-in and password resets. IAM is for workforce and service access, not an end-user directory.

Which AWS Lambda deployment package format supports a function whose unzipped code and dependencies reach roughly 900 MB?

A container image. Zip deployment packages, whether uploaded directly or pulled from Amazon S3, share a 250 MB unzipped limit across the function and its layers. A Lambda container image supports artifacts up to 10 GB, comfortably holding a 900 MB dependency set, and Lambda runs it directly from Amazon ECR as a single artifact.

When two attached Lambda layers ship a file at the same path under /opt, which version does the function load at runtime?

Lambda extracts every attached layer into /opt in the order the layers are listed on the function. Because they share one directory tree, the layer listed later overwrites the earlier layer's file at that same path, and the runtime loads that later copy. Attachment order decides the winner, not layer version numbers or any conflict check.

Common misconceptionThat the platform would refuse a configuration it cannot resolve cleanly. Lambda accepts overlapping layer paths silently and lets the later layer overwrite the earlier one.

To list every recent ERROR log event with its requestId, newest first, directly from a CloudWatch Logs log group, what should you use?

Run a CloudWatch Logs Insights query against the log group: it executes in place over the stored events. Filter on the parsed level field equal to ERROR, project the requestId with a fields clause, and sort by timestamp descending. This returns the exact failing events in time order without exporting data or building a pipeline.

How can a high-throughput AWS Lambda function emit per-request custom CloudWatch metrics with dimensions without a synchronous metrics API call on every invocation?

Emit the log line using the CloudWatch embedded metric format (EMF). EMF encodes metric definitions and dimensions in a metadata section of a structured JSON log event, so one log write captures both the queryable log and the custom metrics. CloudWatch extracts the metrics asynchronously from the logs, avoiding a per-request PutMetricData call.

Common misconceptionThat a PutMetricData call per request is a reasonable way to record custom metrics. At high throughput it adds a synchronous call, latency and cost to every invocation.