Amazon AWS Certified Solutions Architect - Professional SAP-C02 Bundle

WAF rules are deployed without visibility into which requests they affect. The stated evidence instead indicates: Shared costs are assigned by arbitrary percentage without service consumption evidence.

Instances remain oversized based only on purchase history rather than performance evidence. The stated evidence instead indicates: Shared costs are assigned by arbitrary percentage without service consumption evidence.

Teams purchase more RIs independently without measuring existing shared coverage. The stated evidence instead indicates: Shared costs are assigned by arbitrary percentage without service consumption evidence.

Every reporting query was sent to the cluster writer endpoint despite available readers. The stated evidence instead indicates: Single-AZ database storage and compute are a single zonal failure domain.

The function's unqualified ARN always invokes latest code for all users, bypassing version routing. The stated evidence instead indicates: Single-AZ database storage and compute are a single zonal failure domain.

Unbounded request bursts overload backends because no gateway rate protection is configured. The stated evidence instead indicates: Single-AZ database storage and compute are a single zonal failure domain.

Region-specific public IPs and DNS routing do not provide the needed global anycast transport optimization. The stated evidence instead indicates: Stream shard limits or skewed partition keys cannot handle incoming record throughput.

The destination bucket is present but replication cannot use the destination KMS key. The stated evidence instead indicates: Stream shard limits or skewed partition keys cannot handle incoming record throughput.

Instance-local caches cannot share consistent ephemeral state across horizontally scaled hosts. The stated evidence instead indicates: Stream shard limits or skewed partition keys cannot handle incoming record throughput.

Team A reads only its permitted dataset and an attempt to fetch team B's restricted dataset is rejected. However, this does not establish the required outcome: failed deliveries retry under the configured policy and exhausted messages reach the DLQ.

Unhealthy containers leave active target selection promptly while healthy targets continue receiving requests. However, this does not establish the required outcome: failed deliveries retry under the configured policy and exhausted messages reach the DLQ.

The stack tolerates the injected zonal fault and stop alarms terminate experiments as planned. However, this does not establish the required outcome: failed deliveries retry under the configured policy and exhausted messages reach the DLQ.

Threat detection remains inconsistently deployed across member accounts and Regions. The stated evidence instead indicates: Administrators run sensitive maintenance scripts with no enforced approval or trail.

Untracked imperative orchestration repeats previous side effects after transient failures. The stated evidence instead indicates: Administrators run sensitive maintenance scripts with no enforced approval or trail.

Console edits accumulate because no one compares deployed state with infrastructure definitions. The stated evidence instead indicates: Administrators run sensitive maintenance scripts with no enforced approval or trail.

Failed notifications disappear because no supported retry and DLQ design is in place. The specific problem here is that availability claims rest only on architecture diagrams, not controlled chaos evidence.

Only instance host metrics exist and individual container bottlenecks cannot be isolated. The specific problem here is that availability claims rest only on architecture diagrams, not controlled chaos evidence.

Kubernetes deployment treats process startup as adequate evidence of request readiness. The specific problem here is that availability claims rest only on architecture diagrams, not controlled chaos evidence.

The disaster-recovery Region has no local image copy and the deployment assumes remote registry availability. The stated evidence instead indicates: Console-created infrastructure diverged from versioned architecture definitions.

Objects remain indefinitely in higher-cost storage despite a proven low-access retention pattern. The stated evidence instead indicates: Console-created infrastructure diverged from versioned architecture definitions.

Individual account stacks were created manually and drift from the approved baseline. The stated evidence instead indicates: Console-created infrastructure diverged from versioned architecture definitions.

A simulated interruption causes bounded rework and lowers total compute cost without losing final results. However, this does not establish the required outcome: failure of one Region does not interrupt healthy remaining regional capacity under synthetic load.

A restore drill meets the 24-hour recovery target with verified data integrity and access. However, this does not establish the required outcome: failure of one Region does not interrupt healthy remaining regional capacity under synthetic load.

The load balancer continues serving from healthy remaining tasks during a forced zone failure. However, this does not establish the required outcome: failure of one Region does not interrupt healthy remaining regional capacity under synthetic load.

Eligible baseline compute spend receives discounts without excessive underused commitment. However, this does not establish the required outcome: release decisions reference tracked error rate and latency burn rather than raw instance uptime alone.

An incident view identifies the affected service and response latency rather than just total fleet CPU. However, this does not establish the required outcome: release decisions reference tracked error rate and latency burn rather than raw instance uptime alone.

Simulated or published applicable Health events reach the responsible operations group before planned impact. However, this does not establish the required outcome: release decisions reference tracked error rate and latency burn rather than raw instance uptime alone.

A pilot demonstrates the chosen mainframe transformation path and acceptable business results. However, this does not establish the required outcome: destination table state stays synchronized and the application passes a reversible cutover test.

Partners connect over SFTP and transferred objects appear in assigned bucket prefixes. However, this does not establish the required outcome: destination table state stays synchronized and the application passes a reversible cutover test.

Kubernetes manifests and controllers operate within approved permissions and the control plane is managed. However, this does not establish the required outcome: destination table state stays synchronized and the application passes a reversible cutover test.

An all-at-once replacement has unacceptable outage and rollback risk. The stated evidence instead indicates: Console changes have no reproducible declarative infrastructure state.

All consumers wait for a nightly export despite a near-real-time freshness requirement. The stated evidence instead indicates: Console changes have no reproducible declarative infrastructure state.

Dedicated EC2 servers remain the scaling bottleneck for stateless containerized workloads. The stated evidence instead indicates: Console changes have no reproducible declarative infrastructure state.