Amazon AWS Certified Solutions Architect - Associate SAA-C03 Bundle

The option Use Amazon EKS where Kubernetes controllers are required addresses a separate condition: a portable Kubernetes operator controls complex workload custom resources. EKS provides Kubernetes APIs and custom resource extensibility. The actual requirement is small container services need simple scaling without Kubernetes cluster operations. Recommended corrective action: Deploy them on Amazon ECS with Fargate. ECS/Fargate offers managed task scheduling without Kubernetes.

Run the application on EC2 with approved instance configuration would help if a vendor requires full OS access and specialized kernel modules. EC2 exposes guest OS control unavailable in serverless containers. That is not the symptom described here: small container services need simple scaling without Kubernetes cluster operations. The matching response is Deploy them on Amazon ECS with Fargate, because eCS/Fargate offers managed task scheduling without Kubernetes.

For the alternative failure a sustained batch transform needs custom binaries and long runtimes, the control Use AWS Batch or ECS tasks for long-running containerized processing is relevant: batch and ECS support runtimes unsuitable for Lambda. Here the evidence instead is small container services need simple scaling without Kubernetes cluster operations. The selected technical mechanism is Deploy them on Amazon ECS with Fargate: eCS/Fargate offers managed task scheduling without Kubernetes.

Configure suitable CloudWatch metric aggregation and scaling cooldowns would help if app instances frequently scale in and out due to noisy short spikes. Dampening scale changes reduces thrashing. That is not the symptom described here: cPU remains low but requests queue behind a slow downstream API. The matching response is Scale on an appropriate backlog or latency metric rather than CPU alone, because the chosen scaling signal should reflect the bottleneck.

For the alternative failure scale-out hits the EC2 instance quota before adding requested capacity, the control Request supported service quota increases and monitor scaling limits is relevant: quotas can cap effective autoscaling capacity. Here the evidence instead is cPU remains low but requests queue behind a slow downstream API. The selected technical mechanism is Scale on an appropriate backlog or latency metric rather than CPU alone: the chosen scaling signal should reflect the bottleneck.

The competing option Define bootstrapping and instance warmup/health grace behavior would be useful when a new ASG begins accepting traffic before application initialization completes. Startup time must be excluded from premature unhealthy replacement. The present issue is cPU remains low but requests queue behind a slow downstream API; it is more directly addressed by Scale on an appropriate backlog or latency metric rather than CPU alone. The chosen scaling signal should reflect the bottleneck.

Use ALB request-count-per-target or other workload-specific tracking solves a different operational problem where the service uses an ALB but its scaling policy references unrelated network metrics. Request-aware signals correlate better with app demand. It does not directly address app instances frequently scale in and out due to noisy short spikes. The case calls for Configure suitable CloudWatch metric aggregation and scaling cooldowns, since dampening scale changes reduces thrashing.

The competing option Define bootstrapping and instance warmup/health grace behavior would be useful when a new ASG begins accepting traffic before application initialization completes. Startup time must be excluded from premature unhealthy replacement. The present issue is app instances frequently scale in and out due to noisy short spikes; it is more directly addressed by Configure suitable CloudWatch metric aggregation and scaling cooldowns. Dampening scale changes reduces thrashing.

The option Request supported service quota increases and monitor scaling limits addresses a separate condition: scale-out hits the EC2 instance quota before adding requested capacity. Quotas can cap effective autoscaling capacity. The actual requirement is app instances frequently scale in and out due to noisy short spikes. Recommended corrective action: Configure suitable CloudWatch metric aggregation and scaling cooldowns. Dampening scale changes reduces thrashing.

Use ALB request-count-per-target or other workload-specific tracking solves a different operational problem where the service uses an ALB but its scaling policy references unrelated network metrics. Request-aware signals correlate better with app demand. It does not directly address a new ASG begins accepting traffic before application initialization completes. The case calls for Define bootstrapping and instance warmup/health grace behavior, since startup time must be excluded from premature unhealthy replacement.

The competing option Scale on an appropriate backlog or latency metric rather than CPU alone would be useful when cPU remains low but requests queue behind a slow downstream API. The chosen scaling signal should reflect the bottleneck. The present issue is a new ASG begins accepting traffic before application initialization completes; it is more directly addressed by Define bootstrapping and instance warmup/health grace behavior. Startup time must be excluded from premature unhealthy replacement.

The option Request supported service quota increases and monitor scaling limits addresses a separate condition: scale-out hits the EC2 instance quota before adding requested capacity. Quotas can cap effective autoscaling capacity. The actual requirement is a new ASG begins accepting traffic before application initialization completes. Recommended corrective action: Define bootstrapping and instance warmup/health grace behavior. Startup time must be excluded from premature unhealthy replacement.

The option Define bootstrapping and instance warmup/health grace behavior addresses a separate condition: a new ASG begins accepting traffic before application initialization completes. Startup time must be excluded from premature unhealthy replacement. The actual requirement is scale-out hits the EC2 instance quota before adding requested capacity. Recommended corrective action: Request supported service quota increases and monitor scaling limits. Quotas can cap effective autoscaling capacity.

Scale on an appropriate backlog or latency metric rather than CPU alone would help if cPU remains low but requests queue behind a slow downstream API. The chosen scaling signal should reflect the bottleneck. That is not the symptom described here: scale-out hits the EC2 instance quota before adding requested capacity. The matching response is Request supported service quota increases and monitor scaling limits, because quotas can cap effective autoscaling capacity.

For the alternative failure app instances frequently scale in and out due to noisy short spikes, the control Configure suitable CloudWatch metric aggregation and scaling cooldowns is relevant: dampening scale changes reduces thrashing. Here the evidence instead is scale-out hits the EC2 instance quota before adding requested capacity. The selected technical mechanism is Request supported service quota increases and monitor scaling limits: quotas can cap effective autoscaling capacity.

The option Define bootstrapping and instance warmup/health grace behavior addresses a separate condition: a new ASG begins accepting traffic before application initialization completes. Startup time must be excluded from premature unhealthy replacement. The actual requirement is the service uses an ALB but its scaling policy references unrelated network metrics. Recommended corrective action: Use ALB request-count-per-target or other workload-specific tracking. Request-aware signals correlate better with app demand.

Scale on an appropriate backlog or latency metric rather than CPU alone would help if cPU remains low but requests queue behind a slow downstream API. The chosen scaling signal should reflect the bottleneck. That is not the symptom described here: the service uses an ALB but its scaling policy references unrelated network metrics. The matching response is Use ALB request-count-per-target or other workload-specific tracking, because request-aware signals correlate better with app demand.

For the alternative failure app instances frequently scale in and out due to noisy short spikes, the control Configure suitable CloudWatch metric aggregation and scaling cooldowns is relevant: dampening scale changes reduces thrashing. Here the evidence instead is the service uses an ALB but its scaling policy references unrelated network metrics. The selected technical mechanism is Use ALB request-count-per-target or other workload-specific tracking: request-aware signals correlate better with app demand.

Use SQS standard queue with scalable consumers and idempotent handlers would help if a high throughput worker must consume many independent standard messages. Standard queues scale broadly with at-least-once unordered delivery. That is not the symptom described here: messages for each customer must be processed in order while customers process concurrently. The matching response is Use an SQS FIFO queue with customer-specific message-group IDs, because fIFO groups preserve per-group ordering while allowing parallel groups.

For the alternative failure a consumer receives a message but crashes without deleting it, the control Let visibility timeout expire and process idempotently on redelivery is relevant: unacknowledged messages become available for retry. Here the evidence instead is messages for each customer must be processed in order while customers process concurrently. The selected technical mechanism is Use an SQS FIFO queue with customer-specific message-group IDs: fIFO groups preserve per-group ordering while allowing parallel groups.

Redrive failed messages to a configured DLQ after threshold solves a different operational problem where a message repeatedly fails due to invalid business schema. DLQ isolation prevents unbounded poison-message retry. It does not directly address messages for each customer must be processed in order while customers process concurrently. The case calls for Use an SQS FIFO queue with customer-specific message-group IDs, since fIFO groups preserve per-group ordering while allowing parallel groups.

For the alternative failure a message repeatedly fails due to invalid business schema, the control Redrive failed messages to a configured DLQ after threshold is relevant: dLQ isolation prevents unbounded poison-message retry. Here the evidence instead is a producer retries sending the same logical operation and creates duplicate deliveries. The selected technical mechanism is Set FIFO deduplication ID or enable content-based deduplication within its limits: fIFO deduplication can suppress duplicate sends in its window.

Let visibility timeout expire and process idempotently on redelivery solves a different operational problem where a consumer receives a message but crashes without deleting it. Unacknowledged messages become available for retry. It does not directly address a producer retries sending the same logical operation and creates duplicate deliveries. The case calls for Set FIFO deduplication ID or enable content-based deduplication within its limits, since fIFO deduplication can suppress duplicate sends in its window.

The competing option Use SQS standard queue with scalable consumers and idempotent handlers would be useful when a high throughput worker must consume many independent standard messages. Standard queues scale broadly with at-least-once unordered delivery. The present issue is a producer retries sending the same logical operation and creates duplicate deliveries; it is more directly addressed by Set FIFO deduplication ID or enable content-based deduplication within its limits. FIFO deduplication can suppress duplicate sends in its window.

Use SQS standard queue with scalable consumers and idempotent handlers solves a different operational problem where a high throughput worker must consume many independent standard messages. Standard queues scale broadly with at-least-once unordered delivery. It does not directly address a consumer receives a message but crashes without deleting it. The case calls for Let visibility timeout expire and process idempotently on redelivery, since unacknowledged messages become available for retry.

The competing option Use an SQS FIFO queue with customer-specific message-group IDs would be useful when messages for each customer must be processed in order while customers process concurrently. FIFO groups preserve per-group ordering while allowing parallel groups. The present issue is a consumer receives a message but crashes without deleting it; it is more directly addressed by Let visibility timeout expire and process idempotently on redelivery. Unacknowledged messages become available for retry.

The option Redrive failed messages to a configured DLQ after threshold addresses a separate condition: a message repeatedly fails due to invalid business schema. DLQ isolation prevents unbounded poison-message retry. The actual requirement is a consumer receives a message but crashes without deleting it. Recommended corrective action: Let visibility timeout expire and process idempotently on redelivery. Unacknowledged messages become available for retry.

The option Redrive failed messages to a configured DLQ after threshold addresses a separate condition: a message repeatedly fails due to invalid business schema. DLQ isolation prevents unbounded poison-message retry. The actual requirement is a high throughput worker must consume many independent standard messages. Recommended corrective action: Use SQS standard queue with scalable consumers and idempotent handlers. Standard queues scale broadly with at-least-once unordered delivery.

Use an SQS FIFO queue with customer-specific message-group IDs would help if messages for each customer must be processed in order while customers process concurrently. FIFO groups preserve per-group ordering while allowing parallel groups. That is not the symptom described here: a high throughput worker must consume many independent standard messages. The matching response is Use SQS standard queue with scalable consumers and idempotent handlers, because standard queues scale broadly with at-least-once unordered delivery.

For the alternative failure a consumer receives a message but crashes without deleting it, the control Let visibility timeout expire and process idempotently on redelivery is relevant: unacknowledged messages become available for retry. Here the evidence instead is a high throughput worker must consume many independent standard messages. The selected technical mechanism is Use SQS standard queue with scalable consumers and idempotent handlers: standard queues scale broadly with at-least-once unordered delivery.

The option Use an SQS FIFO queue with customer-specific message-group IDs addresses a separate condition: messages for each customer must be processed in order while customers process concurrently. FIFO groups preserve per-group ordering while allowing parallel groups. The actual requirement is a message repeatedly fails due to invalid business schema. Recommended corrective action: Redrive failed messages to a configured DLQ after threshold. DLQ isolation prevents unbounded poison-message retry.

Use SQS standard queue with scalable consumers and idempotent handlers would help if a high throughput worker must consume many independent standard messages. Standard queues scale broadly with at-least-once unordered delivery. That is not the symptom described here: a message repeatedly fails due to invalid business schema. The matching response is Redrive failed messages to a configured DLQ after threshold, because dLQ isolation prevents unbounded poison-message retry.

For the alternative failure a consumer receives a message but crashes without deleting it, the control Let visibility timeout expire and process idempotently on redelivery is relevant: unacknowledged messages become available for retry. Here the evidence instead is a message repeatedly fails due to invalid business schema. The selected technical mechanism is Redrive failed messages to a configured DLQ after threshold: dLQ isolation prevents unbounded poison-message retry.