AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 96
Question 950
An ecommerce RDS database has dozens of unbounded manual snapshots but no clear recovery policy. The business requires a point-in-time recovery window and a limited history of clean restore points. What cost-aware practice fits?
- Turn off all automated backups and keep only the latest live DB writer.
- Configure supported automated backups/PITR with approved retention, and prune unnecessary manual snapshots after review.
- Keep every manually created snapshot forever because they consume no additional stored changed blocks.
- Replace PITR with a read replica alone and claim old deleted rows can always be restored from it.
Explanation
Correct Answer
B:Purposeful recovery-point retention meets RPO and historical coverage without indefinite duplicate storage growth.
Incorrect Answers
A:No historical recovery path remains for corruption or accidental deletion.
C:Manual database snapshots can retain changed data blocks and incur backup storage charges beyond any applicable free allowance. Retaining every snapshot without a recovery policy can accumulate unnecessary storage over time.
D:Replicas can propagate unwanted changes and are not a full historical backup mechanism.
Question 951
A business has a statutory requirement to restore selected financial records in another account and Region. A proposal copies every temporary data-science database backup to three Regions as well. What is more economical?
- Copy all test database snapshots forever because inter-Region replication always has zero cost.
- Keep only local backups of regulated financial records even though the policy requires independent regional recovery.
- Copy critical backups selectively according to the cross-account/Region recovery scope and leave recreatable transient backups out unless justified.
- Replace backup copies with a Cost Explorer tag and assume tags make data recoverable in the destination.
Explanation
Correct Answer
C:Recovery-boundary coverage should prioritize business-critical datasets rather than duplicate all low-value data indiscriminately.
Incorrect Answers
A:Extra copies consume storage and related transfer/operation resources.
B:The recovery requirement explicitly includes a failure of the primary Region. Backups located only in that same Region may be unavailable during a regional outage, so independent-region recovery copies are needed for the stated scope.
D:Cost Explorer tags attribute resource spending for analysis; they do not create, replicate or preserve database recovery points. A tag on a database cannot substitute for a cross-Region backup that can actually be restored.
Question 952
An RDS administrator sees many historical snapshots but assumes that because snapshots are incremental, they are always free after the first one. Finance wants to evaluate long-term backup charges. What is accurate?
- Treat all changed blocks in subsequent manual RDS snapshots as free because the snapshots are incremental.
- Budget only the free automated backup allowance and ignore separately retained manual snapshots and extended history.
- Delete older snapshots solely by creation date without checking whether the last usable historical recovery point or legal retention is lost.
- Measure billed additional/retained changed snapshot blocks, backup allocation and retention before deleting unneeded snapshots.
Explanation
Correct Answer
D:RDS snapshots are incremental for changed backup blocks, but billable storage depends on retained unique blocks, applicable automated-backup allowances and manually retained copies. Measure actual consumption and required recovery history before deletion.
Incorrect Answers
A:Incremental snapshots avoid recopying unchanged blocks but newly changed blocks and retained history still consume charged backup storage beyond eligible allowances.
B:Extra retained manual snapshots and copied backups can exceed included space even when daily automated backups fit an allowance.
C:A retention decision must account for restore dependencies, corruption detection windows and required historical restore states, not age alone.
Question 953
A database team wants to cut backup retention from 30 days to one day to reduce storage cost. It has never measured whether a ten-day-old corruption could be recovered or how long restoring a large backup takes. What should precede the decision?
- Test restores at a range of recovery points, including one older than the proposed retention window, and verify recovery timing and required corruption-detection lookback before reducing retention.
- Keep only the most recent automated snapshot and rely on Multi-AZ failover to restore a clean database before a corruption first detected ten days later.
- Reduce retention to one day and add a larger instance class to shorten restore time, assuming performance capacity replaces missing historical recovery points.
- Use only the latest continuously replicating read replica as the recovery mechanism, even when logical data corruption might replicate to it.
Explanation
Correct Answer
A:A tested retention policy must cover the age of potential undetected corruption and meet required RPO/RTO. Actual restore trials reveal whether historical clean points remain available and can be recovered in time.
Incorrect Answers
B:Multi-AZ improves availability under certain failures but can replicate logical corruption. A one-day backup window would not provide a ten-day-old clean restore point.
C:A larger database class might speed some operations but cannot reconstruct backups that no longer exist. Retention and recovery-point selection must be tested independently of compute size.
D:A read replica can improve read scaling or recovery topology, but asynchronous replication may carry corrupted writes; it is not a substitute for historical backups.
Question 954
A company migrates a commercial relational schema to PostgreSQL. Table copies succeed through a replication task, but stored procedures and vendor-specific SQL functions fail on the target. What separate tool/process is needed?
- Enable the AWS DMS change data capture task on the same source tables and expect the replication engine to rewrite unsupported stored procedure bodies.
- Run AWS Schema Conversion Tool compatibility assessment, convert eligible database objects and manually remediate unsupported procedural SQL alongside DMS or other data movement.
- Create a new AWS Glue Data Catalog database with matching table names, treating the catalog metadata as translated PostgreSQL procedure implementations.
- Use an Amazon RDS read replica of the source engine as the PostgreSQL destination, relying on physical replication to translate proprietary database SQL.
Explanation
Correct Answer
B:AWS SCT and migration assessments can convert supported schemas and code; vendor-specific functions, triggers and application SQL often require manual review. DMS is a separate data-movement component.
Incorrect Answers
A:DMS CDC is used to keep source and target row data synchronized; it does not implement SQL dialect translation or rewrite complex stored procedures.
C:Glue Data Catalog describes metadata and external data tables; creating its database entries does not compile or migrate procedural SQL into PostgreSQL functions.
D:An RDS engine-native read replica is not a heterogeneous engine conversion tool. Moving between incompatible SQL engines requires schema and code transformation plus data migration.
Question 955
A legacy SQL Server application has high recurring license expenses. The platform team considers PostgreSQL on RDS but some features may require rewrite. What should drive the savings decision?
- Use AWS DMS alone to copy SQL Server table data into PostgreSQL and assume it will translate all T-SQL stored procedures and vendor features.
- Estimate PostgreSQL license savings without retiring the existing paid SQL Server fleet or its software entitlements.
- Compare licensing, compute/storage operations and feature/SQL migration effort after validating open-source engine compatibility.
- Move the identical SQL Server stored procedures to PostgreSQL on EC2 without conversion and treat the DB engine switch as only a hosting change.
Explanation
Correct Answer
C:Validate the SQL Server feature inventory and migrate schema/code with suitable assessment and conversion work, then compare license savings to PostgreSQL capacity, ongoing operations, cutover and regression-testing costs.
Incorrect Answers
A:DMS performs migration/replication of supported data, not automatic transformation of every proprietary stored procedure and application query.
B:License savings are only realized to the extent that paid legacy usage and entitlements are actually reduced after verified cutover.
D:Engine differences affect SQL dialect, transaction semantics and extensions; changing the hosting platform does not make proprietary procedures compatible.
Question 956
A migration team successfully cuts over its application's database to a new managed instance but leaves the original large DB running indefinitely for 'just in case' even after the approved rollback window. What is the next cost action?
- Delete the new validated production DB and keep only the old fleet because it has a longer billing history.
- Keep both at full size forever because a completed migration automatically makes the legacy instance free.
- Change only the old DB instance name while preserving all compute and storage allocation.
- Verify cutover and retention obligations, take required final backups, then decommission unused legacy compute/storage safely.
Explanation
Correct Answer
D:Running both complete stacks permanently accrues duplicate charges beyond the defined rollback need.
Incorrect Answers
A:That reverses successful cutover and can lose current transactions.
B:Old running databases continue billing until changed or removed.
C:Changing an RDS DB identifier does not deprovision the underlying database compute, provisioned storage or backups. Charges generally continue until those resources are resized or retired according to the actual configuration.
Question 957
A platform owns RDS and DynamoDB workloads. The monthly bill combines DB instance compute, table requests, backups, storage I/O and data transfers, and managers cannot see what drives growth. Which analysis supports optimization?
- Inspect detailed Cost Explorer/CUR usage categories and isolate billed database compute, storage, requests, I/O and backup lines.
- Assume the entire bill is database instance CPU hours even for DynamoDB request-priced resources.
- Use the number of SQL tables as the entire cost formula for RDS and DynamoDB.
- Disable all cost allocation data because simplifying the invoice makes optimization more precise.
Explanation
Correct Answer
A:Targeted cost reduction requires identifying the usage dimensions rather than blaming a single aggregate DB total.
Incorrect Answers
B:Database pricing models differ and include noncompute categories.
C:Tables alone do not represent billed throughput or retained bytes.
D:Less billing detail makes it harder to identify waste.
Question 958
An enterprise shares databases across departments but wants each product owner accountable for its workload. Query-level data is confidential, so finance prefers supported billing attribution. What should it establish?
- Measure RDS CPU usage by month and distribute all DB expenses according to department headcount without identifying which application owns the resources.
- Use supported tags, linked accounts and appropriate charge allocation dimensions to map RDS/DynamoDB spend to applications.
- Rely only on database resource display names and assume those names become activated cost-allocation tag fields in every billing export.
- Export every confidential SQL statement to the finance team and estimate usage from its text without establishing billing tags.
Explanation
Correct Answer
B:Use supported resource tags, linked-account boundaries and valid billing export dimensions to attribute database spend. Verify tag activation and the service support for each billable resource; any shared capacity needs a documented allocation rule.
Incorrect Answers
A:An aggregate CPU trend and headcount are weak proxies for actual database resource ownership or billable instance/request costs.
C:Names are not a universal substitute for activated resource tags, per-service cost attributes, or linked-account separation.
D:Sharing confidential query content is unnecessary for standard cost allocation and does not reliably map all provisioned DB or request charges.
Question 959
A traffic bug multiplies DynamoDB writes by 40 overnight. The team finds the problem only on next month's bill. Which AWS financial control would provide earlier detection?
- Rely only on annual RI renewal notices to detect a one-night DynamoDB write spike.
- Disable DynamoDB request metrics to reduce alert volume and assume the write spike disappears.
- Set appropriate AWS Budgets and Cost Anomaly Detection notifications scoped to relevant database usage/spend.
- Change the database table name when charges rise and assume prior active requests stop billing.
Explanation
Correct Answer
C:Spend thresholds and anomaly alerts can surface unexpected request growth before a full billing cycle completes.
Incorrect Answers
A:Commitment notices do not watch request-based billing anomalies.
B:Disabling DynamoDB request metrics hides visibility into a surge but does not stop the underlying writes or associated billing. The application must control or remediate the write source while retaining diagnostic telemetry.
D:Naming does not control ongoing service use or pricing.
Question 960
A team has one Amazon RDS for MySQL Reserved DB Instance commitment for a db.r6i.xlarge in us-east-1. Its workload is being split onto two db.r6i.large MySQL instances in the same Region, with the same license model and Single-AZ deployment as before. The finance team asks whether the remaining discount can cover this new shape. Which statement is accurate?
- The reservation must be canceled because splitting one xlarge MySQL instance into two compatible large instances invalidates size flexibility.
- The reservation becomes a Compute Savings Plan and automatically discounts Amazon EC2 instances but not RDS MySQL.
- The reservation discount applies to only one db.r6i.large instance because normalized instance usage cannot be divided between eligible databases.
- The eight normalized units of the eligible db.r6i.xlarge reservation can apply to the two db.r6i.large MySQL instances at four units each.
Explanation
Correct Answer
D:The listed db.r6i classes exist for RDS for MySQL, and one xlarge represents eight normalized units while each large represents four. With the same Region, database engine, class type and compatible deployment, the reservation can cover the two large instances.
Incorrect Answers
A:A change from an xlarge reservation to two smaller eligible RDS for MySQL instances does not by itself invalidate the discount. The reservation can apply across compatible sizes using normalized units.
B:RDS Reserved DB Instances remain a database billing commitment; they do not convert into EC2 Compute Savings Plans or begin covering unrelated compute instances.
C:RDS Reserved DB Instance size flexibility accounts for normalized usage, rather than binding the entire reservation discount to a single physical DB instance. Two large instances together equal one xlarge in this model.
